Diagnosis Related Groups (DRGs) and theMedicare Program: Implications for

Medical Technology

July 1983

NTIS order #PB84-111525

—

Library of Congress Catalog Card Number 83-600560

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

Preface

The increase in the cost of hospital care has been a persistent and growing problemfor both the Medicare program and the general public for more than iS years. A substan-tial portion of the increase in hospital costs has been attributed to an increase in theuse of new and existing medical technologies.

Congress recently legislated a new prospective per-case payment system for theMedicare program. Hospitals will be paid a specific, predetermined amount for eachpatient treated, regardless of the number or types of services provided. The amountpaid will depend primarily on the Diagnosis Related Group (DRG) into which the pa-tient is classified. The implementation of per-case payment has rested on the availabili-ty of an acceptable method of measuring a hospital’s case mix. DRGs are just one ofseveral approaches to measuring hospital case mix. Their importance is heightened bytheir adoption in the new national Medicare prospective payment system.

The House Committee on Energy and Commerce and its Subcommittee on Healthand the Environment requested that OTA examine DRGs and their implications for usein the Medicare program as part of a larger assessment on Medical Technology andCosts of the Medicare Program.

This technical memorandum presents the results of that examination. It reviewsthe development of DRGs and compares them to alternative case-mix measures. It ex-amines the validity and reliability of the DRG classification system and the accuracyof DRG coding. It provides examples of proposed and actual uses of DRGs in hospitalpayment. Finally, and most important, the technical memorandum includes a thoroughanalysis of the implications for medical technology use and adoption of using DRGsas an integral part of a per-case payment system.

A principal finding is that while DRGs are ready for use in a per case paymentsystem, their implementation needs to be closely monitored, because there is little ex-perience with their use in this context. In the long run, the success of DRG paymentwill rest on its flexibility and aptability to changing costs and technologies. Four find-ings concern this need for periodic adjustment:

1. the requirement for regular reestimation of relative DRG prices implies a need forcontinued collection of hospital cost and charge data;

2. procedures allowing for the adjustment of DRG rates conditional on a hospital’sadoption of a technology maybe important to stimulate adoption of desirable butcost-raising technologies;

3. the DRG adjustment process requires supporting evidence about the effectiveness,risks, and costs of new technologies; and

4. the adjustment process should guard against proliferation of DRGs.

Medical Technology and Costs of the Medicare Program Advisory Panel

Stuart Altman, Panel ChairDean, Florence Heller School, Brandeis University

Frank BakerVice PresidentWashington State Hospital Association

Robert BlendonSenior Vice PresidentThe Robert Wood Johnson Foundation

Jerry CromwellPresidentHealth Economics ResearchChestnut Hill, Mass.

Karen DavisJohns Hopkins UniversitySchool of Hygiene and Public Health

Robert DerzonVice PresidentLewin & AssociatesWashington, D. C’.

Howard FrazierCenter for the Analysis of Health PracticesHarvard School of Public Health

Clifton GausCenter for Health Policy StudiesWashington, D.C.

Jack HadleyUrban InstituteWashington, D. C.

Kate IrelandChairman, Board of GovernorsFrontier Nursing ServiceWendover, Ky.

Judith LaveGraduate School of Public HealthUniversity of Pittsburgh

Mary MarshallDelegateVirginia House of Delegates

Walter McNerneyNorthwestern University

Morton MillerPresidentNational Health CouncilNew York, N. Y.

James MonganExecutive DirectorTruman Medical CenterKansas City, Mo.

Seymour PerryInstitute for Health Policy AnalysisGeorgetown University Medical Center

Robert SigmondBlue Cross of Greater Philadelphia

Anne SomersDepartment of Environment and

Community and Family MedicineRutgers University

Paul TorrensUCLA School of Public Health

Keith WeikelGroup Vice PresidentAMIMcLean, Va.

iv

OTA Project Staff—Diagnosis Related Groups and the Medicare Program

H. David Banta, Assistant Director, OTAHealth and Life Sciences Division

Clyde J. Behney, Health Program Manager

Anne Kesselman Bums, Project Director,Medical Technology and Costs of the Medicare Program

Judith L. Wagner, Senior Analyst

Cynthia P. King, Analyst

Pamela Simerly, Research Assistant

Virginia Cwalina, Administrative AssistantJennifer Nelson, Secretary

Mary Walls, Secretary

Other Contributing Staff

Gloria Ruby, Analyst

Contractors

OTA Publishing Staff

John C. Holmes, Publishing Officer

John Bergling Kathie S. Boss Debra M. Datcher Joe Henson

Glenda Lawing Linda Leahy Donna Young

v

Contents

Page

Glossary of Acronyms and Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

vii

Contents (continued)

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79

List of Tables

Table No. PageI. Malpractice Claims Paid, 1978 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282. Ratesetting Regression Coefficients, Selected Facilities and Services . . . . . . . . . . . . . . . . 383. Four Types of Technological Innovations and Effects on Capital Cost,

Operating Cost, and Total Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404. Impact of Technological Innovation on Per-Case Costs . . . . . . . . . . . . . . . . . . . . . . . . . . 42B-1. Summary of Similarities and Differences Between the Original and

Modified Digs ....,.... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

viii

Glossary of Acronyms and Terms

Glossary of Acronyms

AHA — American Hospital AssociationCPD – cost per dischargeCPHA — Commission on Professional and

Hospital ActivitiesCT – computed tomographyDHHS — Department of Health and Human

ServicesDRGs — Diagnosis Related GroupsEEG – electroencephalogramGIR — Guaranteed Inpatient RevenueHCFA — Health Care Financing AdministrationIOM — Institute of MedicineI P P B – intermittent positive pressure breathingLos — length of stayMDCS — Major Diagnostic CategoriesNCHS — National Center for Health StatisticsOTA — Office of Technology AssessmentPAS — Professional Activity StudyPCB – preliminary cost basePMCS — Patient Management CategoriesPROS — Peer Review OrganizationsPSROS — Professional Standards Review

OrganizationsSMSA — Standard Metropolitan Statistical AreaTEFRA — Tax Equity and Fiscal Responsibility

Act of 1982

Glossary of Terms

Capital costs: Costs associated with the use of capitalfacilities and equipment, including depreciation andinterest expenses.

Case mix: The relative frequency of admissions ofvarious types of patients, reflecting different needsfor hospital resources. There are many ways ofmeasuring case mix, some based on patients’ diag-noses or the severity of their illnesses, some on theutilization of services, and some on the characteris-tics of the hospital or area in which it is located.

Coinsurance: A form of cost-sharing whereby the in-sured pays a percentage of total cost. (Also see co-payment. )

Copayment: A form of cost-sharing whereby the in-sured pays a specific amount at the point of con-sumption, e.g., $10 per visit. (Also see coinsurance. )

Cost-sharing: The general set of financing ar-rangements whereby the consumer must pay someout-of-pocket cost to receive care, either at the timeof initiation of care, or during the time of the pro-vision of health care services, or both.

Deductible: A form of cost-sharing in which the in-sured incurs an initial expense of a specified amountwithin a given time period (e.g., $250 per year)before the insurer assumes liability for any addi-tional costs of covered services.

Diagnosis Related Groups (DRGs): A classificationsystem that groups patients according to principaldiagnosis, presence of a surgical procedure, age,presence or absence of significant comorbidities orcomplications, and other relevant criteria.

DRG inflation: An increase over time in the numberof separately identified case-mix classificationgroups.

DRG payment: Any per-case hospital payment meth-od in which differences in case mix are taken intoaccount using DRGs to classify case types.

Effectiveness: Same as efficacy (see below) except thatit refers to “ . . . average or actual conditions ofuse. ”

Efficacy: The probability of benefit to individuals ina defined population from a medical technology ap-plied for a given medical problem under ideal con-ditions of use.

Fee-for-service: A method of paying for medical careon a retrospective basis by which each service ac-tually received by an individual bears a relatedcharge.

Length of stay (LOS): The number of days a patientremains in the hospital from admission to discharge.

Medicaid: A Federal program that is administered andoperated individually by each participating Stategovernment that provides medical benefits to cer-tain low-income persons in need of health and med-ical care.

Medical technology: The drugs, devices, and medicaland surgical procedures used in medical care, andthe organizational and supportive systems withinwhich such care is provided.

Medicare: A nationwide, federally administered healthinsurance program, authorized in 1965 to cover thecost of hospitalization, medical care, and some re-lated services for eligible persons over age 65, per-sons receiving Social Security Disability Insurancepayments for 2 years, and persons with end-stagerenal disease. Medicare consists of two separate butcoordinated programs–hospital insurance (part A)and supplementary medical insurance (part B).Health insurance protection is available to insuredpersons without regard to income.

Pass-throughs: In a per-case payment system, pass-throughs are elements of hospital cost that are paidon the basis of cost-based reimbursement.

Per-case payment: Any prospective hospital paymentsystem with fixed rates of payment based on the hos-

98-823 0 - 83 - 2ix

—

pital admission, not on the number and types ofservices or number of days of care provided.

Professional Standards Review Organizations(PSROs): Community-based, physician-directed,nonprofit agencies established under the Social Se-curity Amendments of 1972 to monitor the qualityand appropriateness of institutional health care pro-vided to Medicare and Medicaid beneficiaries.

Prospective payment: Hospital payment programswhere rates are set prior to the period during whichthey apply and where the hospital incurs at leastsome financial risk.

Recalibration: Periodic changes in relative DRG prices,including assignment of prices to new DRGs.

Reliability: A measure of the consistency of a methodin producing results. A reliable test gives the sameresults when applied more than once under the sameconditions. Also called “precision. ”

Retrospective cost-based reimbursement: A paymentmethod in which hospitals are paid their incurredcosts of treating patients after the treatment has oc-curred.

Risk: A measure of the probability of an adverse oruntoward outcome and the severity of the resultantharm to health of individuals in a defined popula-tion associated with use of a medical technology ap-plied for a given medical problem under specifiedconditions of use.

Safety: A judgment of the acceptability of risk (seeabove) in a specified situation.

Technology: The application of organized knowledgeto practical ends.

Technology assessment: A comprehensive form ofpolicy research that examines the technical, eco-nomic, and social consequences of technological ap-plications. It is especially concerned with unin-tended, indirect, or delayed social impacts. In healthpolicy, the term has also come to mean any formof policy analysis concerned with medical technol-ogy, especially the evaluation of efficacy and safe-ty. The comprehensive form of technology assess-ment is then termed “comprehensive technology as-sessment .“

Utilization and quality control peer review organiza-tions (PROS): Physician organizations establishedby the Tax Equity and Fiscal Responsibility Act of1982 (Public Law 97-248) to replace PSROs (see defi-nition). Hospitals are mandated to contract withPROS to review quality of care and appropriatenessof admissions and readmissions.

Validity: A measure of the extent to which an observedsituation reflects the “true” situation. Internal validi-ty is a measure of the extent to which study resultsreflect the true relationship of a “risk factor” (e.g.,treatment or technology) to the outcome of interestin study subjects. External validity is a measure ofthe extent to which study results can be generalizedbeyond the study sample.

Chapter 1

Introduction and Summary

Chapter 1

Introduction and Summary

BACKGROUND

The increase in the cost of hospital care has beena persistent and growing problem for both theMedicare program and the general public for morethan 15 years. A substantial portion of the in-crease in hospital costs has been attributed to anincrease in the use of new and existing medicaltechnologies. * Medicare expenditures for inpatienthospital services have increased more than ten-fold since its inception—from about $3 billion in1967 to more than $33 billion in 1982. From 1979to 1982, the average cost of a day of hospital careincreased at an annual rate of almost 18 percent,and Medicare expenditures for hospital servicesincreased at a rate of over 19 percent. In 1982,hospital costs increased by 15.5 percent, threetimes the rate of inflation in the economy as awhole.

While the fiscal health of the program suffersas a result of hospital cost inflation, Medicare hascontributed to the problem through its traditionalhospital payment policy. Until October 1982,Medicare employed a retrospective cost-based re-imbursement approach whereby hospitals couldrecover from Medicare most of what they spentfor Medicare beneficiaries. Consequently, hos-pitals had little incentive to control costs. Hos-pitals have thus been encouraged to acquire anduse more technology and to expand their capaci-ty to produce a wider scope of more complexservices.

The Medicare program has historically pro-vided leadership for other hospital payment pol-icies. Other third-party payers, including StateMedicaid programs and private insurers, havealso generally used cost-based reimbursement astheir approach to hospital payment. In fact, until1981, State Medicaid programs were required tofollow Medicare’s principles of reimbursement for

● OTA has defined medical technology as the drugs, devices, andmedical and surgical procedures used in medical care, and the or-ganizational and supportive systems within which such care is de-livered. In this technical memorandum, the focus is on drugs, devices,and procedures, but many of the points apply to the system tech-nologies.

hospitals unless they applied for and received awaiver from the Federal Government for an alter-native system.

However, as early as the late 1960’s, some ofthese other payers began the search for alter-natives to retrospective cost-based reimburse-ment. State-legislated and voluntary programsusing alternative payment schemes have appearedthroughout the 1970’s. These programs have beenbroadly termed “prospective payment,” whererates are set prior to the period during which theyapply and the hospital incurs at least some finan-cial risk. They have varied widely in design. Forexample, some control the amounts hospitalscould charge for specific services; others payhospitals an all-inclusive rate per day of hospitalcare. But paying by the day sets up incentives forhospitals to increase lengths of stay and admis-sions, and controlled charges also encourage hos-pitals to increase the number of services theyprovide.

Recently, a new kind of prospective paymenthas emerged: per-case payment. Under this formof payment, the hospital is paid a specific amountfor each patient treated, regardless of the numberor types of services provided. Thus, the hospitalis rewarded for reducing the cost of treating a pa-tient over the entire course of the hospital stay.Per-case payment removes the incentive to pro-vide more technologies and encourages the hos-pital and its physicians to consider explicitly thebenefits of additional services against their addedcosts.

Per-case payment cannot survive for long with-out a method to adjust for differences in the kindsof patients that hospitals treat. If hospitals werepaid the same amount for each admission regard-less of its clinical characteristics, over time theywould be encouraged to treat patients who areless ill and to avoid the cases that require moreresources. Thus, the implementation of per-casepayment has rested on the availability of an ac-ceptable method of measuring the hospital’s casemix.

3

4

Case mix has been defined in various ways. Inthis technical memorandum, it refers to the rel-ative frequency of admissions of various types ofpatients, reflecting different needs for hospitalresources. There are many ways of measuring casemix, some based on patients’ diagnoses or theseverity of their illnesses, some on the utilizationof services, and some on the characteristics of thehospital or area in which it is located.

Diagnosis Related Groups (DRGs) are just oneof several approaches to measuring hospital casemix. Their importance is heightened by their re-cent adoption for use in the new national Med-icare prospective payment system. Beginning inOctober 1983, Medicare will phase in a per-casepayment system using DRGs as the case-mixmeasure. As part of a larger assessment on Medical Technology and Costs of the Medicare Pro-gram, the Office of Technology Assessment(OTA) was requested by the House Committeeon Energy and Commerce and its Subcommitteeon Health and the Environment to examine DRGsand their implications for use in the Medicareprogram.

This technical memorandum, Diagnosis RektedGroups (DRGs) and the Medicare Program: Im-plications for Medical Technology presents theresults of that examination. As with all OTA tech-nical memoranda, it contains no policy optionsfor congressional consideration. It is intended tobe a comprehensive and independent assessmentof DRGs in the context of a per-case payment sys-

SUMMARY

Alternative Case-Mix Measures

An examination of several case-mix measuresfor their validity and acceptability in a per-casepayment system reveals DRGs to be the best avail-able measure. The Disease Staging and the Severi-ty of Illness Index methods of measuring case mixprovide more information about the severity ofthe condition of the patients, but both requiremore data than are generally available and bothare based on subjective methods. Neither measurehas reached the stage of development where it issuitable for widespread implementation in a pay-

tem. In its study of DRGs as a case-mix measure,it reviews their development and compares themto alternative case-mix measures. It examines thevalidity and reliability of the DRG classificationsystem, the accuracy of DRG coding, and the ad-ministrative feasibility of administering a DRG-based payment system. It provides examples ofproposed and actual uses of DRGs in hospital pay-merit. ’ Finally, the technical memorandum in-cludes a thorough analysis of the implications formedical technology use and adoption of usingDRGs as an integral part of a per-case paymentsystem. This analysis includes a review of the keyfeatures of design of DRG payment systems thataffect medical technology and a discussion of theimplications of technological change for the ad-ministration of a DRG payment system over time.

Two issues of concern to policymakers are notincluded in this technical memorandum. First, itdoes not address the effect of DRG payment onthe costs of the Medicare program; rather, itdiscusses the incentives that will be established bysuch a system. Second, it does not discuss whetherand to what extent DRG payment under Medicarewill lead to savings in Medicare program costs atthe expense of other payers. These critical issuesare presently under debate by other agencies andorganizations.

● A detailed account of one actual use—the experimental use ofDRGs for hospital payment in New Jersey-is presented in a separateworking paper.

ment context. Nevertheless,alternative approaches does

the existence of thesereinforce the concern

of some health providers and policymakers re-garding the adequacy of DRGs in distinguishingdifferences in the relative severity of patients’ con-ditions in any given DRG.

Another case-mix measure, Patient Manage-ment Categories (PMCS), is also in the develop-mental stage and will be tested soon. PMCs dif-fer from other case-mix measures, includingDRGs, in that they are normative. Physiciansspecify an optimal set of clinical care components

5

based on a patient’s clinical characteristics, in-cluding both final diagnosis and reason for ad-mission. This set of clinical care components isthe basis for the relative cost weights of PMCs.This system appears unique in that it recognizesthat optimal patient management should be thefocus of a system that seeks to encourage efficien-cy. Thus, the further development of PMCsshould be encouraged.

The use of DRGs in the Medicare per-case pay-ment system is appropriate since they are morerefined than the alternative case-mix measures.Both statistical and clinical considerations supportthis conclusion. Since DRGs can be assigned basedon the information already processed on the dis-charge abstracts of patients’ medical records, itis superior to the other measures in its admin-istrative feasibility. However, empirical evidencemust still be collected on all of the alternativemeasures to compare them in the context of pay-ment.

DRG Payment and the Use ofMedical Technology

There are two general incentives inherent in anyper-case payment system: 1) to reduce the costto the hospital of each inpatient case stay, and2) to increase the number of inpatient admissions.Cost per case can be reduced by using fewer tech-nological services, including ancillary services,reducing the number of inpatient days, or both.This incentive may result in specialization amonghospitals for services that require a minimumnumber of patients to maintain profitability. Thisspecialization may imply lower access to care forsome Medicare patients. There are built-in con-straints of unknown magnitude on the possibili-ty of adverse effects on access and quality. Oneconstraint is the fact that physicians are the deci-sionmakers, and they continue to have financial,ethical, and legal reasons to practice high-qualitymedicine.

The direction and strength of general incentivesfor any particular hospital are altered by keyfeatures of the DRG payment system, including:1) the proportion of the hospital’s case loadcovered by DRG payment, 2) the treatment of

costs as pass-throughs, * 3) the methods of DRGrate construction, 4) the methods of updatingDRG rates, and 5) the level of risk and rewardbuilt into the payment system. Thus, a DRG pay-ment system may include a variety of specific ap-proaches to payment with some predictable ef-fects on medical technology.

DRG payment incentives may be expected toaffect technology use in the following ways:

Overall, the number and intensity of an-cillary procedures provided to inpatients canbe expected to decrease, but the use of pro-cedures that can be shown to lower the costper case will increase.The settings of technology use are likely tobe influenced by DRG payment, but the in-centives work in conflicting directions andare sensitive to the key features of programdesign. It remains to be seen which incentivewill dominate for which procedures. DRGpayment will encourage the movement oftechnologies into the home, particularly thosefor post-hospital care.DRG payment is likely to influence the spe-cialization of services, but the magnitude anddirection of these effects is unknown. The in-centives to reduce costs encourage concen-tration of capital-intensive technologies infewer institutions. Conversely, the increas-ing competition among hospitals for physi-cians and patients will create incentives forthe widespread acquisition of some technol-ogies.A change in technology product mix is like-ly to result from downward pressure on theprice and quantity of supplies and, if capitalis included in the DRG rate, capital equip-ment. Greater product standardization canbe expected as more expensive models andprocedures are eased out of the marketthrough competition.

*“Pass-throughs” are elements of hospital cost that are not con-trolled by the per-case payment system. Cost-based reimbursement,as a whole, can be interpreted as a payment method in which allcost categories are passed through.

6

Effects of DRG Payment onTechnological Change in Medicine

Perhaps even more important than how DRGpayment affects the use of presently availablemedical technologies is how DRG payment willaffect technological change in medicine—theadoption of new technologies and discarding ofold ones. DRG payment will influence hospitals’decisions to adopt new medical technologies andmay therefore alter the rate and direction of tech-nological change in medicine.

Although no empirical studies on the effect ofDRG payment on adoption of technologies areavailable, studies of other kinds of prospectivepayment systems suggest that hospitals can anddo respond to changes in financial incentives inthese decisions. In general, technologies that arecost-reducing to hospitals will be encouraged;cost-raising technologies will be discouraged.However, much depends on the strength and de-sign of the program. In particular, the methodsof providing rewards for cost reductions, treatingcapital costs, and updating DRG prices have im-portant implications for technological change.Though DRG payment does not imply that tech-nological change will approach a standstill, itsdirections are likely to be altered, and the adop-tion of technologies that are cost-raising to thehospital is likely to decline by an unknown quan-tity.

The longrun viability of any DRG payment sys-tem depends on its ability to both adapt to andencourage appropriate technological change inmedicine. The methods and procedures used toadjust the average payment level, relative DRGrates, and the DRG categories themselves are crit-ical to the survival of the system. The objectivesof the adjustment process are to maintain equali-ty across DRGs in the ratios of price to cost ofefficient care and to promote the adoption of ap-propriate new technologies. There are at least fivepotential processes of adjustment. They vary ac-cording to whether the adjustment is conditionalon hospitals’ actual adoption of new technology,who requests the adjustment, and when in thestage of a new technology’s diffusion the adjust-ment is made. None of these processes alone is

sufficient to adjust the system adequately for tech-nological change.

Implementation Issues

Other considerations for the feasibility of usingDRGs as the case-mix measure in the Medicarepayment system include two important aspectsof implementation of this new system: 1) data andcoding issues, and 2) hospital administrative is-sues. Accurate and timely patient-level data areclearly important to the efficient and effectiveoperation of the DRG system. In the past, evalua-tions of patient discharge data have found themto be unreliable. However, it is important to notethat these abstracts had not been produced forpayment purposes. When payment depends onthe accuracy and timeliness of discharge abstracts,their importance increases, and data reliabilityshould improve. Monitoring by utilization andquality control peer review organizations as man-dated by the new Medicare law, should give hos-pitals added incentive to improve their data col-lection and coding.

Such improvements in information quality im-plies a need for several education programs formedical staff, hospital administrators, and med-ical records personnel. Error detection, feedback,and training would be important parts of dataprograms. It should be noted that these types ofimprovements are likely to be more expensive andtime-consuming. Some of these costs will varyamong the individual hospitals depending on theircurrent practices. For example, some hospitalsmight need to adopt computer capability for med-ical records, while others might need to add totheir medical records staff.

Findings and Conclusions

● Although the new Medicare law provides fora DRG-based per-case payment system, DRGshave been inadequately evaluated for theirvalidity as an indicator of patient resourceneeds and for their impact on medical technol-ogy in per-case payment. Thus, it is critical thatthe new system’s implementation be carefullymonitored.

7

●

●

●

●

Programs of quality assurance and utilizationreview will be required to counter the incen-tives of the per-case payment system to under-provide services and inappropriately admit anddischarge patients.

The treatment of capital costs will affect the useof medical technology. The diversity in hos-pitals’ ages, debt structures, and future needsfor expansion and closure argue for hospital-specific determinations of capital payment lev-els, probably at the State level.

Periodic reestimation of relative DRG rates toreflect changes in the costs of various DRGs isessential to a workable program. This impliesa need for a continuing source of cost andcharge data to support the process.

Methods for updating DRG rates that are con-ditional on technology adoption maybe impor-tant to stimulate desirable but cost-raising tech-nologies, Frequent creation of new technology-specific DRGs, however, can ultimately under-mine the incentives of per-case payment.

●

●

●

The DRG adjustment process requires support-ing evidence about the effectiveness, risks, andcosts of new technologies. Resources must beadequate for necessary research and the ac-tivities of groups such as the Prospective Pay-ment Assessment Commission.

It is fortunate that the new Medicare law doesnot discourage individual States from establish-ing alternative prospective payment systems.These alternative systems will allow experimen-tation with different payment system configura-tions, including the use of other case-mix meas-ures as they become more refined.

The reliance of the DRG classification systemon accurate and timely data collection and cod-ing will necessitate improvement of hospitals’medical records procedures and performance.Educational programs for physicians, nurses,hospital administrators, and medical recordspersonnel could be initiated.

ORGANIZATION OF THE TECHNICAL MEMORANDUM

This technical memorandum is organized in sixchapters. Chapter 2 reviews and evaluates severalalternative approaches to case-mix measurementfor payment purposes. In addition to DRGs, Dis-ease Staging, Severity of Illness Index, and PMCsare briefly examined. Chapter 3 analyzes the in-centives for hospitals to use medical technologiesunder a prospective per-case payment systembased on DRGs. The effects of DRG payment ontechnological change in medicine is the focus ofchapter 4. Implementation issues regarding dataand administration are briefly described in chapter5. Chapter 6 provides an overview and expansionof the conclusions reached in previous chapters.

Appendix A includes a list of the Health Pro-gram Advisory Committee and acknowledgmentsof assistance in the preparation of this technical

memorandum. Appendix B provides a descriptiveoverview of the development of DRGs, and ap-pendix C contains brief descriptions of per-casepayment systems that have been designed by theStates.

A separate working paper is entitled “UsingDiagnosis Related Groups (DRGs) in HospitalPayment: The New Jersey Experience” was writ-ten by Joanne Finley under contract to OTA. Itis a detailed description of the New Jersey ex-perience with an all-payer prospective paymentsystem based on DRGs. *

*Available from the National Technical Information Service(NTIS).

98-823 0 - 83 - 3

Chapter 2

A Brief Review and Evaluation ofAlternative Approaches to

Case - Mix Measurement

Chapter 2

A Brief Review and Evaluation ofAlternative Approaches to

Case-Mix Measurement*

INTRODUCTION

Policymakers, hospital administrators, andhealth services researchers have long recognizedthe diversity of hospitals’ outputs. Efforts toanalyze hospital behavior and to establish effec-tive and equitable reimbursement systems havebeen complicated by this diversity of hospitalproducts, which include education and researchas well as patient care.

For two decades, researchers have grappledwith the measurement of hospitals’ patient careoutput. While enormous progress has been madeduring that period, there remains no consensusabout the most appropriate method of case-mixmeasurement. Failure to reach such a consensusis not surprising. Several substantial barriers havestood in the way, including the variety of pur-poses case-mix measures have been designed toserve and the significant data requirements asso-ciated with any but the most rudimentary meas-ures. The variation in the purposes to be servedhas been a barrier, because it appears that the “op-timal” measure may be different if it is to be usedfor reimbursement, quality assurance, manage-ment, or for some other purpose. Data require-

EARLY APPROACHES

The earliest measures of patient care outputwere developed at the institutional level. That is,the initial measures represented one or more in-dices, developed for the hospital as a whole, thatwere designed to reflect a dimension of hospitalperformance assumed to be associated with casemix. These measures included average length ofstay (LOS), surgery rates, relative volume of out-patient visits, number of births, and other similar

● This chapter is based on a paper prepared for OTA by NancyL. Kelly, Diane E. Hamilton, and Ralph E. Berry of Policy Anal-ysis, Inc.

ments stem from the need, irrespective of thespecific measure, to obtain detailed informationabout the patients for which the case-mix measureis to be defined. Even in the smallest hospitals,admissions total about 2,000 annually; in thelargest institutions, more than 40,000 patients areadmitted each year. Even in this age of computers,the national number of admissions—3l million*—is formidable.

This chapter presents a brief overview of thedevelopment of case-mix measures, from the earlyrudimentary techniques to the most recent ad-vances. It is intended to provide a frame ofreference against which Diagnosis Related Groups(DRGs) can be assessed. It is not intended to pro-vide a detailed review of past or current ap-proaches to case-mix measurement. Such detailcan be found in the references cited throughoutthe chapter or, alternatively, in an excellentreview article by Hornbrook (38).

— . — —“Total admissions to non-Federal short-term general hospitals in

1980.

indicators. Data to construct such measures werereadily available from published sources, such asthe American Hospital Association’s annual sur-vey of hospitals.

It was recognized that simple summary indicessuch as these were no better than crude surrogatesfor case mix. However, including such measuresin an analysis of interhospital cost variation, forexample, seemed preferable to excluding any out-put characteristics, and for a number of years

11

12

there were no alternatives. But the empirical evi-dence indicated that these measures did not ex-plain very much of the interhospital variationin costs (see, e.g., 44). Clearly, the low ex-planatory power resulted in part from the unidi-mensionality of the measures. The ratio of surgicaloperations to admissions, for example, fails to dis-tinguish hospitals performing a great deal of com-plicated, high-risk surgery from those perform-ing equal numbers of simpler, more common pro-cedures. Similarly, a long LOS may be experi-enced by patients with acute and severe illnessesor by patients with chronic conditions awaitingdischarge to a nursing home.

Often, additional surrogate measures were usedalong with those described above. The addedmeasures described characteristics of the hospital,rather than the patient population, but they wereconsidered to be highly correlated with case mix.The earliest surrogate was the absolute size of thehospital, measured in terms of numbers of bedsor admissions. Case mix was assumed to be morecomplex in larger hospitals than in smaller fa-cilities. Teaching status is another commonly usedsurrogate; it is widely believed that teaching hos-pitals treat a more severely ill mix of patients thannonteaching hospitals. Similarly, physician staffcharacteristics have occasionally been identifiedas useful surrogates. Measures of the scope offacilities and services have also been used as in-dicators of the case mix of patients treated, underthe assumption that hospitals are equipped to treata particular array of illnesses. While size, teachingstatus, physician staff composition, and scope ofservices undoubtedly have some predictive powerwith respect to case mix, the evidence suggests thatthis power is far from perfect (see, e.g., 4s).

A later development among the early ap-proaches involved the linkage of demographic andeconomic characteristics of the hospital’s “marketarea” (from sources such as the U.S. Bureau ofCensus) to the hospital-specific information. Thiswas based on the further assumption that unmeas-ured dimensions of case mix, not contained in theinstitutional measures, could be obtained from thecharacteristics of the area in which the hospitalwas located. Examples of the characteristicsthought to be important were the age distributionof the population (especially the proportion over

age 65), median income and education levels, andthe numbers of Medicaid recipients. Even such in-direct measures as urban versus rural locationand/or population density were considered. As-sociated with each of these characteristics was anunderlying hypothesis about its effect on the casemix of area hospitals. For example, a high pro-portion of poor and/or elderly people in the sur-rounding area was thought to be more closely as-sociated with severe illnesses than a high propor-tion of well-educated, moderate income house-holds (70). Broad descriptors of the area, such aspopulation density, have often been consideredas surrogates for differences in lifestyle, which inturn, lead to differences in health status.

While it can be argued that they add impor-tant information to the limited hospital-specificindices, the market area characteristics also mustbe viewed as particularly crude surrogates for casemix. The chief drawback of these measures liesin their imprecision. Given current data sources,it is not possible to accurately identify, on a na-tional scale, the precise market area from whichthe hospital draws its patients. This is especiallydifficult in urban areas, where many hospitals areclustered in a small geographic area, and wheresome of the hospitals are referral centers for amuch larger region. Except for the relatively fewStates in which patient origin data are collected,counties or Standard Metropolitan StatisticalAreas are typically the only geographic units forwhich data can be obtained. At best, these willbe rough approximations of the true market areasfor most of the Nation’s hospitals.

It should be noted that the measures that havebeen labeled “early” are, in fact, still widely inuse in research on hospital costs and behavior to-day. Although, as mentioned previously, consid-erable advances have been made in the area ofcase-mix measurement, none of the more recentdevelopments is yet widely available, nor are therequisite input data. As the next section describes,the data requirements of the principal alternativeshave placed new demands on traditional record-keeping and data collection procedures. Gradual-ly, however, alternative measures are likely to be-come more widely available as existing systemsrespond to these demands.

13

RECENT DEVELOPMENTS: DESCRIPTION AND EVALUATIONOF NEW APPROACHES

Dimensions of Case Mix

Recognition that case-mix measurement was animportant problem—and probably a necessarytool for developing solutions to such longstandingdilemmas as hospital cost inflation-aroused con-siderable interest as well as funding for a numberof research and development efforts. As a result,several important advances have been made. Be-fore they are described, however, a conceptualframework for viewing the problem of case-mixmeasurement will be presented as background.

Perhaps the first “advance” that motivated thenew developments was the recognition that thepatient, not the institution, was the appropriateunit of analysis. No matter what the hospital wasequipped to do, where it was located, or whoserved on its physician staff, the types of patientsit treated during the course of a given year werethe true determinants of the hospital’s patient careoutput. The “condition” of the patients in thatpopulation was believed to dictate treatment pat-terns and, consequently, resource use within thehospital. Further, it is useful to consider “condi-tion” as having two key dimensions: the natureof theproldem underlying hospitalization (usuallyindicated by the diagnosis or diagnoses), and therelative severity of that problem.

A major difficulty facing evaluators of case-mixmeasures lies in choosing the appropriate frameof reference for such an evaluation. For instance,some measures, such as DRGs, were developedspecifically (though not exclusively) to accountfor differences in resource use. Other approachesdo not explicitly address resource use, althoughthey may in fact be highly correlated with thisindicator. An evaluation of how well a givenmeasure explains variation in resource use, there-fore, should in fairness recognize the purpose thatthe measure was designed to serve (though thismay not alter the conclusions reached).

Related to this issue are the different perspec-tives of potential users of case-mix measures andtheir implications for the validity and acceptabilityof a given measure. For example, measures that

seem intuitively reasonable, exhibit high ex-planatory power in statistical analysis, and canbe constructed from readily available data sourcesmay be perfectly acceptable to officials of reim-bursement agencies, but they may have no mean-ing or legitimacy for clinicians. Conversely, meas-ures that are acceptable to clinicians may be in-feasible to employ in a large-scale, national pro-gram. It seems unlikely that a single measure willsatisfy the requirements of all potential users,though some approaches will come closer thanothers. The likelihood that different conclusionswill be reached by different groups of users is notnecessarily a problem, however, since alternateapproaches may productively be used in tandem.

Against this background, two broad categoriesof approaches will be described. The first consistsof institutional measures, which (to distinguishthem from the early, traditional approaches) arederived from data describing the diagnostic com-position of the hospital’s patients. Included in thiscategory are the “ad hoc” grouping methods (45),the Resource Need Index (100), and informationtheory measures (17,33). The second categoryconsists of patient-level measures. Of these, themost well known is DRGs. The other approachesreviewed in this chapter include two major effortsto measure severity of illness within a diseaseentity: Disease Staging (24) and the Severity ofIllness Index (30,31). Finally, the discussion in-cludes a new approach to defining patient cate-gories and assessing treatment patterns, knownas Patient Management Categories (PMCs) (107).These last three approaches should be viewed aspotential alternatives to DRGs, although theSeverity of Illness Index can also be viewed as acomplement to them.

Hospital-Level Measures

The institutional-level measures referred toabove all represent aggregations of patient datadesigned to capture the overall case-mix complex-ity of, and the resource implications for, a par-ticular hospital. Of these, the most rudimentarycan be termed the “ad hoc” grouping techniques.

14

These involve the collection of diagnostic infor-mation for patients from a sample of hospitals andaggregating those data in ways that appear to beanalytically meaningful. For example, Lave andLave (45) used this approach to define a set ofcharacteristics for each hospital they studied. Intheir study, a set consisted of proportions of pa-tients in broad diagnostic categories, such as car-diovascular diseases. The categories selected wereassumed to distinguish groups of patients thathave different resource requirements, although nodata were available to directly evaluate that hy-pothesis.

A measure that takes resource use directly intoaccount is the Resource Need Index (RNI), devel-oped by the Commission on Professional andHospital Activities (100). The construction of RNIis a two-step process. The first step involves thedevelopment of “relative need units” for eachdiagnosis (or diagnostic category). The units,which are based entirely on charge information,represent the ratio of average charges for a par-ticular diagnosis to overall average charges. RNIis then constructed as the average number of rel-ative need units for a given hospital. RNI thusmakes a start at the simultaneous evaluation ofdiagnosis and resource use, though a possibledrawback lies in its reliance on charges as the soleindicators of resource requirements. To the ex-tent that differences between charges and costs dif-fer by type of service, hospitals’ charges may notreflect actual resource use very precisely.

Finally, information theory measures have alsobeen derived for the institution as a whole. Studiesby Evans and Walker (17) and Horn and Schu-macher (33) have employed this approach to case-mix measurement. Information theory measuresare based on the assumptions that rare conditionsare complex and tend to be treated by a few spe-cialized hospitals. In contrast, common conditionsare assumed to be less complex and likely to beuniformly distributed across hospitals. The meas-ure is derived mathematically from the propor-tions of cases in each diagnostic category. Thehighest scores for this measure will therefore goto the “most complex” hospitals—i.e., those thattreat the uncommon diseases. The specific re-source requirements of any of the diseases, in-cluding the rare ones, are not explicitly considered

in the information theory approach. Instead, itsvalidity in measuring case-mix complexity restson the validity of underlying assumptions.

The only common threads in these divergenttechniques are their use of diagnostic data andtheir assumptions about the appropriateness ofaggregating the information to the hospital level.While this aggregation is undoubtedly convenient,it involves considerable simplification and, as aconsequence, loss of information. Some, such asKlastorin and Watts (44), have considered theissue of hospital-level case-mix indices and haveconcluded that summary indices may not be com-parable across hospitals. However, the thrust ofmost recent research in the area of case-mixmeasurement has been toward the developmentof patient-level measures. Though it is possibleto aggregate these measures into a hospitalwideindex (and indeed several such indices have al-ready been constructed), such aggregation has notbeen a principal focus of the development process.

Patient-Level Measures

Diagnosis Related Groups

DRGs are undoubtedly the most well knownof the patient classification systems that have beenintroduced during recent years. As this section in-dicates, however, several major alternatives nowexist that differ from DRGs both conceptually andin practice. All of the prominent systems are de-scribed briefly in this chapter. A more completedescription of DRG development methods can befound in appendix B.

The development of DRGs has been ongoingsince the late 1960’s, and it is appropriate to viewthe concept as one that is continuously evolving.The evolution of DRGs has involved both con-ceptual refinements and technical improvements,spurred by the availability of more and betterquality input data and by feedback from a widevariety of observers and users of DRGs. It is likelythat the evolution will continue as relevant dataincrease in availability and improve in quality andas the concept is subjected to more and more scru-tiny.

The first version of DRGs to be widely dissem-inated was a set of 383 categories, described by

their developers in 1980 (19). Subsequently, inearly 1982, a second and much revised set of 467categories was issued (103). This revised set borelittle resemblance to the “original” 383, as it wasbased on different definitional procedures and adifferent coding convention. Both sets had severalcommon objectives. Both were designed to iden-tify patients with similar expected resource use,measured by length of hospital stay. Both versionswere defined so as to be medically meaningful tophysicians, the key decisionmakers within thehospital with respect to patient care, though theoperationalization of this objective varied sig-nificantly between the two. Finally, both sets ofDRGs were deliberately based on data that arecommonly available, and both sets sought to belimited to “manageable” numbers of groups.

In general, the broad outlines for the construc-tion of both sets of DRGs were the same for eachversion. Actual patient stays in a sample of hos-pitals were the units of analysis. Each patient’sprincipal diagnosis—i.e., the principal reason(after study) for his/her hospitalization—wascoded using a detailed coding system that allowedfor many thousands of possibilities. The first step,therefore, was to collapse the detailed diagnosiscodes into meaningful, but broad, subcategoriescalled “Major Diagnostic Categories (MDCS). ”These MDCs were then further subdivided, usinga combination of statistical analysis and medicaljudgment, according to other characteristics thataccounted for differences in resource use withineach MDC. (The remainder of this discussion willfocus on the current set; a comparison of the twocan be found in app. B.)

The major differences, however, appear to out-weigh the similarities. Significantly modified pro-cedures were used to develop the 467 DRGs.These differences included the involvement of afar greater number of participants, many of themclinicians, which accompanied a shift in the fun-damental orientation of the development process.Whereas the development of the 383 DRGs hadinvolved both statistical analysis and expertclinical judgment, the balance between the twocomponents was relatively more even than it be-came in the revised method, in which the balancewas shifted in favor of clinical judgment.

The current set of 467 DRGs was derived from23 MDCs, most of which were defined aroundorgan systems of the body (in conformance withthe organization of medical practice). SubdividingMDCs into DRGs was performed by expert panelscomprising physicians and others in the health in-dustry. Their decisions were guided by severalcriteria that had been established. For example,the guidelines required that the initial partition(when possible) be based on the presence or ab-sence of a surgical procedure performed in an op-erating room. Panels were also instructed to rankorder surgical procedures according to resourceintensity and to assign patients with multiple pro-cedures to the procedure involving the greatestintensity. In addition to considering the type ofsurgery performed, the nature of coexisting con-ditions and complications were explicitly evalu-ated. “Substantial” conditions and complicationswere distinguished from those less significant. Sur-gery, coexisting conditions, and complicationswere all viewed as indicators of severity of illness.Finally, other variables were taken into accountwhen the experts determined that they were rele-vant. Examples of other factors used for subdivi-sion include death and “left hospital againstmedical advice. ”

Though clinical judgments dominated the deci-sionmaking process, statistical analysis was usedto aid those judgments. Patient-level data weremade available by several organizations, prin-cipally the Professional Activity Study of theCommission for Professional and Hospital Activ-ities. These data were viewed to be representa-tive of national treatment patterns. Reduction invariance for LOS was examined for possible par-titioning variables, but the fact that variance mayhave been significantly reduced by a particularvariable did not guarantee that that variablewould be included in the DRG definition. Theneed to group patients with clinically relateddiseases, above all, dictated which measureswould be used.

The outcome of this process was a mutually ex-clusive and exhaustive set of 467 DRGs. * By de-

*In operation, there are 468 DRGs, the last for patients who havereceived an operating room procedure unrelated to their MDC.

98-823 0 - 83 - 4

16

sign, DRGs can be determined from discharge ab-stract data, which are commonly available incomputerized form. A computer algorithm isavailable to classify each patient into the ap-propriate DRG. As a consequence, use of thisclassification system poses few administrativeburdens. As the following chapter will describe,little empirical research has been conducted todate on the current set of DRGs. Some recent evi-dence indicates that there still remains substan-tial within-DRG variation in resource use. For ex-ample, one analysis of a random sample of casesin 12 high-volume DRGs applicable to older pa-tients in New Jersey hospitals found that 13.6 per-cent of discharges had a LOS more than two timesthe average of cases in that category (75). Analysisof Medicare discharges for 1979 also showed awide range of within-DRG variation around themean cost per discharge. For both the old 383DRG and the new 467 DRG classification systems,it was found that DRGs had coefficients of varia-tion ranging from about 0.2 to 1.5 in the Medicarepopulation (93). A coefficient of variation of 0.2can be interpreted as indicating that roughly two-thirds of patients in the DRG have costs within20 percent above or below the mean cost of theDRG. A coefficient of 1.5 indicates that abouttwo-thirds of patients in the DRG lie within ISO

percent of the mean cost. The new DRGs do notappear to increase the homogeneity of the groupswith respect to their actual consumption of re-sources. Finally, the extent to which DRGs areacceptable to clinicians is unclear.

New Jersey and several other States have usedDRGs in hospital payment systems with varyinglydegrees of success. Appendix C contains briefdescriptions of several State per-case paymentsystems, some of which used DRGs.

Disease Staging

Apart from DRGs, the most prominent patient-level measure in the literature is Disease Staging.Both Disease Staging and the Severity of IllnessIndex (to be described subsequently) were devel-oped to address the perceived need to measure theseverity of the patient’s illness as well as his orher diagnosis. Severity has been defined as the riskof death or permanent impairment resulting fromthe illness (38).

Staging consists of the specification of pro-gressive levels of severity for disease in terms ofthe events and pathophysiological observationsthat characterize each stage (24). As described inHornbrook (38), staging involves the segmenta-tion of a disease entity into five primary stages,which are defined as follows:

● Stage O: No disease present, or diagnosisunknown.

● Stage 1: Disease is certain and no complica-tions are present either local or systemic.

● Stage 2: Disease process is limited to an organor system; significantly increased risk of com-plications.

● Stage 3: Significantly greater problem thanstage 2: multiple site involvement; general-ized systemic involvement; poor prognosis.

● Stage 4: Death or most severe stage possible(i.e., the final event of the illness).

Six, rather than five, primary stages are used forcancers, to maintain consistency with previouswork. (In fact, staging was first used in oncologyin clinical trials of new treatments to incorporateillness severity into experimental design and evalu-ation (38). ) Substages have also been defined forthe cancers.

Stage assignments can be made by a computeralgorithm, based on data recorded on dischargeabstracts. Computer-assigned stages, however,may represent underestimates of the stages thattechnicians would derive manually from medicalrecords. Again, according to Hornbrook (38), un-derestimates may occur for two reasons: 1) theprimary diagnostic coding systems used for dis-charge abstracts are not sufficiently precise, and2) insufficient data are included on the dischargeabstracts.

Staging is the product of physicians’ judgmentsabout the biological progression of a givendisease. First and foremost, it was developed tobe a clinically meaningful concept. The extent towhich costs, charges, or LOS varied within stageswas not considered during the development proc-ess. Although some limited evaluation has indi-cated that the stages of a disease are indeed sys-tematically related to differences in those othermeasures (6,23), the relationship between thestages of a disease and resource consumption has

17

not yet been investigated in depth. Further, it hasnot been demonstrated that the individual stagesare homogeneous with respect to resource use(23). In addition, the stages are not necessarilycomparable across diseases, as each disease enti-ty is staged separately. Consequently, stage 2 ofa surgical condition may be much more seriousthan stage 2 of a medical condition and thus re-quire more resources during treatment.

The primary advantages of Disease Staging liein its apparent meaningfulness to clinicians, aswell as in the direct way in which the stages cap-ture the biological severity of illness within a givendiagnosis. Staging has the added advantage of re-quiring only information commonly available oncomputerized discharge abstract data (althoughsome precision is sacrificed when computerizedmethods are used). A significant drawback lies inthe likelihood that certain diseases cannot bestaged. (In general, medical conditions are moredifficult to stage than surgical.) The fact thatresource consumption was not an explicit consid-eration in developing the stages (and as a resultmay or may not be captured by them) may bea serious drawback if stages were used in a reim-bursement context. Moreover, since stages arebased on single diseases and on the prognosis foreach patient, they ignore concurrent conditionsand other patient characteristics which affectresource use, such as social, economic, and psy-chological factors (38).

Severity of Illness Index

Still more recent than staging is a measure thatwas developed to reflect the overall severity ofillness of the patient, and not just the severity ofthe principal diagnosis. The Severity of Illness In-dex, developed at Johns Hopkins University (31),classifies patients into four severity levels. Unlikethe staging procedure, this Severity of Illness In-dex is not disease-specific but instead was designedto apply to all conditions treated in the medicine,surgery, obstetrics/gynecology, and pediatrics de-partments of hospitals. Developed in conjunctionwith a panel of physicians and nurses, the indexis built from seven criteria deemed to be the bestindicators of overall illness severity. Theseinclude:

●

●

●

●

●

●

●

stage of the principal diagnosis;complications of the principal condition;concurrent, interacting conditions that affectthe course of hospital treatment;dependency on the hospital staff;extent of nonoperating room procedures;rate of response to therapy, or rate of recov-ery; andimpairment remaining after therapy for theacute aspect of the hospitalization.

Data relevant to each of the above criteria areobtained from the patient’s medical record. Ab-straction of data is performed manually, by atrained rater. Based on the combined pattern ofseverity levels within each of the seven criteria,the rater makes a judgment about the overallseverity of the patient’s condition. The overall in-dex can range from 1 (least severe) to 4 (mostsevere).

The Severity of Illness Index may be used asan adjunct to other patient classification systems,such as DRGs. In that context, refined categoriesof severity can be developed within categories ofpatients. Current research, however, suggests thata preferred use of the Severity of Illness Indexwould be in conjunction with a very broad clas-sification system, such as the 23 MDCsdescribedearlier (30).

The major advantage of the Severity of IllnessIndex, particularly for payment purposes, wouldappear to be the extent to which it explains varia-tion in resource use. In a comparative analysis in-volving six disease conditions, Horn and col-leagues (36) found that the Severity of Illness In-dex produced groups that were more homogene-ous than DRGs, Disease Staging, or PMCS (to bediscussed subsequently). Homogeneity was as-sessed with respect to total charges, LOS, lab-oratory charges, routine charges, and oftenradiology charges. The index has also been shownto be a good predictor of resource use (35). Thisexplanatory power may, in part, result from themethod used to ascribe a severity level to a givenpatient. Although the seven criteria do not ex-plicitly address resource use, some of the criteria(e.g., extent of nonoperating room procedures)are clearly correlated with it, and there may be

18

a tendency on the part of raters to take it into ac-count when forming a judgment about overall se-verity (38).

A drawback of the Severity of Illness Index isits reliance on manual abstraction of data fromthe medical record and on the judgments of in-dividual raters. Although recordkeeping systemscould be modified in such a way that the data nec-essary to construct the Severity of Illness Indexwould be available on computerized discharge ab-stracts, in general, discharge abstract systems arecurrently inadequate for this purpose. Thus, as-signment of severity levels by this method is rel-atively costly. Also, so long as subjective judg-ments are employed in assigning the index values,there are likely to be problems with the reliabili-ty and acceptability of the measure.

The acceptability of the Severity of Illness ofIndex to clinicians is currently unclear. Althoughthe development of the index was in conjunctionwith physicians and nurses, there is as yet no in-dication of how meaningful the index is to clini-cians around the country. At present, the chiefadvantage of this approach seems to be its suc-cess in accounting statistically for variation inresource use.

Patient Management Categories

A criticism that has been leveled at all case-mixmeasures based on discharge diagnosis is that thediagnosis at discharge is not the only relevantdiagnostic information (106). Instead, it is arguedthat diagnosis at the time of admission determinesthe course of treatment that the physician will em-ploy. In other words, not only the diagnosis ofthe patient, but also the reason for admission, willaffect the ultimate LOS and total costs/charges.Reasons underlying admission can range from ob-servation to chemotherapy to major surgery, allof which have vastly different implications forresource use. Young and colleagues (106) alsohave argued for the development of a measure-ment system that avoids building in actual treat-ment patterns, regardless of their appropriateness.They favor a method that is more normative—i.e., one that views patient characteristics andmanagement without regard to current treatmentpatterns that may result from discretionary deci-

sions, differences in the availability of particularfacilities and services, inefficiencies, etc.

As a consequence, Young and colleagues (105)at Blue Cross of Western Pennsylvania have de-veloped an alternative patient classification sys-tem, whereby patients are classified into PMCs *PMCsdiffer from most of the other systems, in-cluding DRGs, in that they are based primarilyon the patient’s clinical characteristics. The defini-tions of PMCsdo not hinge on how the patientwas treated while in the hospital. PMCsdiffer inother ways as well, as will be discussed subse-quently.

Like the other case-mix measures (all to vary-ing degrees), PMCshave been developed in con-sultation with physicians so that they representclinically meaningful entities. Both final diagnosisand reason for admission are consideredsimultaneously in defining the categories. PMCshave been designed to take levels of severity intoaccount, again from a clinical perspective. Foreach PMC, physicians have specified componentsof care (i.e., diagnostic services, treatment pro-cedures, and expected LOS) that, in their view,are required for the effective management of thetypical patient. Thus, a “patient managementpath” has been associated with each PMC. Thesecomponents of care then form the basis for thederivation of relative cost weights for each PMC.Weights are based on actual cost data from sixparticipating hospitals that have been adjustedand allocated to the components of care. The iden-tification of patient management paths and rel-ative costs during the development process isanother distinguishing feature of the PMC ap-proach.

Currently, PMCs are still being defined, al-though it is anticipated that the process will becompleted by the end of the summer of 1983.Computer software is also being developed thatwill enable the automated mapping of patientsinto PMCs. Although discharge abstract data typ-ically do not include information on reason foradmission, Hornbrook (38) reports that prelim-inary research indicates that valid mapping into

● Much of the substance of this discussion was drawn from un-published documents provided by Wanda W. Young (ICM,IOS).

19

PMCS may be possible without collecting addi-tional data.

The principal advantage of PMCs over eitherDRGs or staging would appear to derive from thejoint consideration, during the development proc-ess, of payment and patient management. Whilethe system was designed for use in a payment con-text, actual patterns of use (as has been noted)were not directly considered in defining the cat-egories. However, the relative cost of each PMCis calculated as part of the development process.This can presumably be incorporated into a pay-ment procedure. What is most unique to this sys-tem, however, is its recognition that patient man-

SUMMARY

Until recent years, case mix has been measuredby hospital-level surrogates for patient care out-put. These measures have been derived fromreadily available sources of data and generallyrepresent crude volume and performance meas-ures along with relevant characteristics of thehospital (e.g., teaching status) and location.

The early measures have been useful in explain-ing some of the observed interhospital variations,but it is apparent that these measures do not con-tain the amount of information necessary to ac-curately capture interhospital case mix differences.As a result, considerable effort has been devotedin the past decade to developing new and im-proved measures of patient care output. The mostwell known of these advances, DRGs, are the sub-ject of this technical memorandum. Other ad-vances, including both substitutes for and com-plements to DRGs, have also been reviewed brief-ly in this chapter.

The major advances in measuring health careoutput have been in the area of severity of illnessmeasurement. Disease Staging and the Severityof Illness Index were both designed to provide aframework for classifying diseases according tothe relative severity of the patient’s condition.Both have required extensive developmental workand testing, which are still underway. Use of eithermeasure would require more data than are gen-

agement should be the focus of any system thatseeks to encourage efficiency and the deliberateattempt on the part of the developers to producea system that would simultaneously be meaningfulto physicians and facilitate efficiency improve-ments in the management of patient care.

Because the system is not yet completed, it ispremature to make comparisons between it andother alternatives. For the same reason, no em-pirical evaluations have yet been performed.Clearly, however, PMCs represent an interestinginnovation in the area of case-mix measurementthat has considerable potential.

erally available at the present time, though thestaging approach can be employed using data thatare normally included in hospitals’ computerizedrecords. Neither measure has reached the pointwhere it is suitable for widespread implementa-tion in a reimbursement context. However, theexistence of such measures, at a minimum, servesas a reminder that the relative severity of patients’illnesses is important to consider when measur-ing case mix.

Patient Management Categories represent thenewest of the major advances in case-mix meas-urement. It will still be some time before thesystem is fully operational and adequate testingcan be performed. In the short term, however,this method of patient classification again rep-resents a reminder that currently used methodsmay not be appropriately targeted.

This review of alternative methods of measur-ing hospital case mix has revealed that earlymethods are lacking in precision and that the newapproaches (with the exception of DRGs) are notyet ready for widespread use. CIearly, any “re-fined” system that tackles patient-level case-mixinformation will require considerably more datathan has been employed (or even available) in thepast. Feasibility considerations, therefore, shouldinclude the relative administrative burdensassociated with each measure as well as the stage

20

of development each measure has reached. Most statistical evidence) in order to assess the ap-importantly, the conceptual differences among the propriateness of each for the purposes it is toalternatives should be evaluated (as well as the serve.

Chapter 3

DRG Payment and the Useof Medical Technology

Chapter 3

DRG Payment and the Useof Medical Technology

INTRODUCTION

The use of Diagnosis Related Groups (DRGs)in hospital payment has grown from an experi-ment in a handful of hospitals to national Med-icare policy in just 3 years. At the time of passageof the 1983 Social Security Amendments (PublicLaw 98-21), which established a national Medicareprospective payment system using DRGs, theMedicare program was planning to use DRGs toimplement the hospital expenditure control pro-visions of the Tax Equity and Fiscal ResponsibilityAct of 1982 (TEFRA). Before that, DRG paymenthad been used for only 2 years in 26 hospitals inNew Jersey and even fewer in Maryland. *

*See app. C for detailed descriptions of selected DRG paymentsystems.

DRG PAYMENT: AN OVERVIEW

Theoretically, DRGs could be used in any hos-pital payment method, including retrospectivecost-based reimbursement, but their importancein payment comes from their use as part of pro-spective per-case payment systems. Per-case pay-ment refers to any prospective hospital paymentsystem with fixed rates of payment based on thehospital admission, not on the number and typesof services or number of days of care provided.Per-case payment is a radical departure from tra-ditional cost-based reimbursement and even fromother kinds of prospective payment. One of theunique features of a per-case payment system isthat it cannot survive for long without a way toadjust payment for differences in case mix; other-wise serious inequities among hospitals would belikely to develop, and selective admitting strat-egies would be encouraged.

DRGs represent only one possible approach tocharacterizing hospital case mix; but as the

The rapid acceptance of DRG payment in theabsence of much experience argues for a carefullook at its implications, both good and bad, formedical technology. DRG payment methods es-tablish incentives for the use of medical technol-ogies both within and outside of hospitals thatdiffer markedly from those of retrospective cost-based reimbursement and other kinds of prospec-tive payment. These new incentives have implica-tions for the efficiency and quality of caredelivered to Medicare beneficiaries. This chapterwill examine the implications of DRG paymentfor the amount, characteristics, prices, and set-tings of medical technology use.

previous chapters demonstrate, the DRG systemis the only explicit case-mix measurement ap-proach that is now ready for use in a paymentsystem. It is not surprising, then, that the searchfor a case-mix adjuster has led to DRGs. DRGpayment is defined here as any per-case hospitalpayment method in which differences in case mixare taken into account using DRGs to classify casetypes. Appendix C provides examples of varioustypes of per-case payment methods that have beenapplied or suggested for use by third-party payers.

Per-case payment is possible without the useof DRGs, but any such method must somehowadjust for case mix, if only implicitly. One fre-quently used approach to per-case payment thatdoes not use DRGs or any other explicit case-mixmeasure is to tie each hospital’s future rate percase to its own costs per case in a fixed base year(2). The base year cost is presumably a reflection

98-823 0 - 83 - 523

2 4

of the mix of cases treated by the hospital in thatyear. So long as it can be assumed that the hos-pital’s case mix is stable and not subject tomanipulation, this is a reasonable, though im-precise, implicit case-mix adjustment method. Astime passes, however, the assumptions of stabilityand nonmanipulation of case mix become moreand more tenuous, requiring ever more cumber-some appeals processes or revisions than in asystem with explicit case-mix adjustment. *

There are two general approaches to the use ofDRGs in per-case payment: 1) DRG-specific pricesper case; and 2) a single rate per case that reflectsthe hospital’s case mix determined by a DRG-based case-mix category. The first approach,DRG-specific prices per case, requires the payerto issue a separate rate for each DRG. The payermay pay a unique rate for each DRG in each hos-pital, the same amount for each DRG regardlessof the hospital in which care is rendered, or dif-ferent amounts for any given DRG depending onthe location or particular characteristics of thehospital.

The case-mix category approach requires thathospitals be classified into a number of groups

● Other criticisms can be lodged against this implicit case-mix ad-justment method. The most important is that it rewards past ineffi-ciency. Hospitals which have been relatively efficient in treating pa-tients will have a lower cost base than those which have been lessso prior to initiation of the per-case system. This criticism can alsobe made of some DRG payment systems, particularly those in whichrate per DRG is based totally or partially on the hospital’s own costper DRG.

based on similarities in their case mixes. Allhospitals in a group would be paid a uniform rateper case. A DRG case-mix index is a categoriza-tion method in which each hospital is assigned aunique index value reflecting the relative resourcerequirements of its particular patients. The indexvalue is determined by a formula using DRGs.

The two approaches to DRG payment–DRG-specific prices and DRG-based case-mix adjust-ers—do not differ much from one another. Theprincipal difference is in the time period on whichcase-mix measurement is based. A DRG indexmust be constructed on the basis of case mix insome prior time period, perhaps the most recent-ly completed fiscal year. A DRG-specific pricingsystem adjusts for changes in case mix as they oc-cur. Thus, any fluctuations in case mix that oc-cur either by chance or by a hospital’s actions,such as the introduction of a new service, wouldbe reflected immediately in a DRG-specific pric-ing system but would enter a DRG case-mix ad-justment system only as time passes.

Both kinds of DRG payment—DRG prices andDRG case-mix indexes—have two essential com-ponents: the average level of payment per case;and the relative weights applied to each DRG. Theaverage payment level determines how stringentor generous the payment system is as a whole,while the relative DRG weights or prices deter-mine the profitability of one DRG relative toanother. The financial incentives of a DRG pay-ment system depend on both the average level andthe relative weights.

DRG PAYMENT AND THE USE OF MEDICAL TECHNOLOGY

Appendix C describes eight per-case paymentsystems, five of which use DRGs. Three DRGpayment systems have already been implemented,and one was recently enacted for the Medicareprogram. This section presents an analysis of theexpected effects of per-case payment, and spe-cifically DRG payment, on access to and use ofmedical technologies. For the purposes of thistechnical memorandum, medical technology isdefined as the drugs, devices, and medical andsurgical procedures used in medical care and the

organizational and supportive systems withinwhich such care is provided. In this technicalmemorandum, the focus is on drugs, devices, andprocedures, but many of the points apply to thesystem technologies.

DRG payment establishes a new set of finan-cial incentives for hospital behavior that differsfrom those found under both cost-based reim-bursement and other kinds of prospective hospitalpayment. These incentives are rooted in per-case

25

payment itself, but their effects differ with the par-ticular case-mix measure adopted. For example,while financial incentives facing hospitals aregenerally the same under the old 383 DRGs andthe new 467 DRGs, their strength and the abilityof hospitals to respond to them may differ. Thisdiscussion concentrates on the new DRGs on theassumption that they are more refined than theold DRGs and, at present, are more practical thanany alternative case-mix measurement approach(see ch. 2). It should be understood that the basicincentives are the result of paying by the case andwill remain to some extent regardless of the case-mix measurement approach taken.

Despite the fact that DRG payment has beenembraced by Congress and the administration inthe past 9 months, there is no empirical evidenceavailable on its effect on access to or use of med-ical technologies. Evaluations of New Jersey’s andMaryland’s DRG systems on the use of serviceswithin or outside of hospitals are not yet avail-able. These programs are themselves so new orof such limited scope that they cannot offer em-pirical evidence on which to draw conclusions.

Evidence does exist on the effect of other typesof prospective hospital payment on the use ofmedical technologies. As part of a comprehensivestudy of nine State-legislated hospital ratesettingsystems, Worthington and Piro (102) found thatprograms that pay hospitals on the basis of a per-diem rate all produced an increase in hospitals’average lengths of stay (LOS) and occupancyrates. This result would be expected from a per-diem ratesetting system in which the longer pa-tients stay, the more revenue the hospital receives.However, a per-diem ratesetting program shouldalso encourage increases in rates of inpatient ad-mission, but no such admission effects werefound. These findings suggest that manipulatingadmission rates may be more difficult than in-creasing the length of hospital stay for thosealready admitted. Taken as a whole, however, theresults do suggest that decisionmakers in hospitalsrespond in predictable ways to financial incentivesfor the use of hospital services. Consequently, inthe absence of empirical evidence on the effectsof DRG payment on medical technology use, an

assessment of the direction and strength of itsfinancial incentives is reasonable.

General Incentives of DRG Payment

To understand how DRG payment affects in-centives to use particular medical technologies,it is helpful first to examine incentives affectingthe use of hospital and other health services ingeneral. These general incentives ultimately trans-late into specific demands for medical technolo-gies.

DRG payment creates two fundamental incen-tives: to reduce the cost to the hospital of eachinpatient hospital stay; and to increase the numberof inpatient admissions.

Incentives To Reduce Cost Per Case

The incentive to reduce cost per case is themotivation for per-case payment in the first place.Per-case payment is predicated on the belief thathospitals have many opportunities to save moneyby operating more efficiently and offering a morecost-effective mix of services, Per-case paymentrewards hospitals that take advantage of these op-portunities.

Reductions in cost per admission can beachieved by reducing LOS, the number or mix ofservices provided during the stay, or the pricespaid for inputs into the production of hospitalservices. Reductions in LOS are likely to have thegreatest immediate effects on per-case costs, al-though such savings would be lower for hospitalsalready operating at low occupancy rates. A re-duction in occupancy rate does not result in a pro-portional reduction in operating costs, becausemany of these (e.g., utilities, housekeeping, ad-ministration) may be largely fixed. Thus, in hos-pitals with low occupancy, the incentive to reduceLOS, though present, will be less than in hospitalswith a high daily census and a backlog of poten-tial admissions. Recent studies have demonstratedthat the well-known regional differences in aver-age LOS in the United States persist even whendiagnosis and severity of illness are taken into ac-count (37,89). Thus, there may be substantialroom for reduction of LOS in some areas of thecountry.

26

Shorter LOS could have positive or negativeeffects on patients’ health. * On the one hand, hos-pitalization itself carries the risk of iatrogenic ill-ness; shorter lengths of stay reduce this risk. Psy-chological factors associated with hospitalizationmay also be important in adversely affecting out-comes. On the other hand, too early dischargecould place patients at risk of inadequate care andthreaten recovery. For example, patients withserious infections have often remained hospital-ized simply to receive long-term intravenous an-tibiotic therapy. There is suggestive evidence thathospital-sponsored home antibiotic therapy pro-grams can save total hospital costs and be safe ifaccompanied by adequate patient training andmonitoring (43,82). However, the potential forinadequate education and followup by hospitalpersonnel exists. While financial incentives underDRG payment would encourage home intrave-nous antibiotic therapy, they would also discour-age the expenditure of resources to educate andmonitor patients adequately.

The incentives inherent in DRG payment re-garding the use of particular ancillary services arecomplex. The cost of ancillary services whose usewould, on the average, shorten hospital LOS,would be weighed against the savings from reduc-tions in LOS. The effect on any particular an-cillary service would depend on the nature of thesecost tradeoffs. For example, hospitals might pro-vide more high-cost antibiotics prophylacticallyif these were shown to substantially reduce theaverage LOS through reductions in hospital-ac-quired infection rates. Or, as another example,liaison psychiatric services, which appear toshorten LOS of postoperative elderly patients (47),might be provided more frequently under DRGpayment than under cost-based reimbursement.A probable byproduct of DRG payment will bean increase in the demand for and supply of in-formation on such cost tradeoffs. Nevertheless,if the consensus is correct that ancillary services,particularly diagnostic tests, have been providedin the past without adequate consideration fortheir impact on total hospital costs (1,26,52,56)then the net effect of per-case payment would be

*This topic is the subject of OTA’s Health Technology Case Study#24, “Variations in Hospital Length of Stay: Their Relationship toHealth Outcomes. ”

to reduce the intensity or amount of these servicesper stay.

The incentive to reduce the price of technologiessuch as drugs and medical supplies is obvious. Inthe past 10 years, hospitals have increasingly em-braced membership in group purchasing plans andgeneric substitution programs. For example, hos-pital membership in pharmacy purchasing groupsgrew from 40 to 88 percent between 1975 and 1981(16,83). Generic substitution–the automatic sub-stitution of a less costly but chemically equivalentgeneric drug for a prescribed brand-name drug—has become commonplace in U.S. hospitals: 96percent of hospitals responding to a nationalsurvey in 1981 reported having such programs(16). The pressure to find new ways to save onthe purchase of drugs and supplies should con-tinue. A logical outcome of this trend is a declinein product variation as hospitals and their pur-chasing groups seek further price reductions andstrengthen the competitive position of productswith high sales volumes.

Incentives To Increase Admissions

DRG payment encourages hospitals to increaseadmissions selectively. Whereas cost- and charge-based reimbursement gave the hospital an incen-tive to keep occupancy rates high by increasingeither admissions or LOS, only admissions pro-duce or increase revenue under DRG payment.Every new admission generates new revenue (inthe amount of the DRG price) and new costs.Serving patients in some DRGs will be more prof-itable than in others, because those DRGs willhave higher ratios of price to cost. The hospitalwould naturally want to encourage the more prof-itable admissions. If the average level of paymentis high enough that all DRGs are profitable, thenthe hospital has an incentive to increase admis-sions in general, but the most profitable admis-sions should still be sought more vigorously.

A variety of mechanisms is available to increaseadmissions selectively, including recruitment ofphysicians in key specialties, adoption of servicesuseful in certain DRGs, and marketing campaignstargeted to preferred patients or their physicians.These strategies may be called “competitive” inthat they are designed to draw patients from otherhospitals.

27

As competition for admissions increases underper-case payment, some specialization in servicedelivery can be expected (5). Since the per-unitcosts of major services often decline as servicevolumes increase, hospitals with high service vol-umes in specific DRGs will find them more prof-itable, and those with low volumes less. Whena hospital finds that a service is unprofitable andwhen the prospects for more efficient operationor increases in volume are dim, it may abandonthe service. For example, a hospital in New Jerseyrecently closed its hyperbaric chamber because itwas found to be unprofitable under DRG pay-ment. Those in need of hyperbaric services (pri-marily divers) will be referred to a hospital in NewYork City (64). However, competition for admis-sions can also drive hospitals to maintain unprof-itable services if their existence is important to themaintenance of the hospital’s position with physi-cians or patients.

Specialization in service delivery may have de-sirable effects on quality as well as cost, since formany services there is a positive relationship be-tween quality and volume (5). However, thesegains in quality and cost could be partially ortotally offset by reductions in patient access toservices. Since it is difficult to predict the kindsof services that will be subject to specializationunder DRG payment, the desirability of futurepatterns of service availability is unknown. AsDRG payment is implemented nationwide, pat-terns of service specialization among hospitalsshould be carefully monitored.

Hospitals may turn to noncompetitive strategiesto increase admissions and lower per-case costs.For example, physicians or staff might be en-couraged directly or indirectly to hospitalizemarginally ill patients and to discharge and re-admit patients at a later date for deferrable pro-cedures that might otherwise be performed as partof a single stay. This “revolving door” incentiveis a new phenomenon, unique to per-case pay-ment. For example, a patient under treatment forpneumonia might be found during the course ofthe hospital stay to have a urological conditionrequiring a deferrable therapeutic procedure.Rather than initiate therapy during the first stay,the physician might discharge the patient for re-admission at a later date. This strategy is both

easy for physicians to implement and difficult forthird-party payers to control.

The incentive to increase admissions selective-ly has its counterpart in an incentive to avoid ad-mitting unprofitable patients. Patient selectionstrategies could conceivably be used to excludepatients in unprofitable DRGs or unprofitable pa-tients within a DRG. But there are important re-strictions on the potential for direct manipulationof case loads. Although hospitals may be able toavoid admissions in some unprofitable DRGs bynot offering the necessary facilities or services, formany patients the DRG is unknown at the timeof admission. Moreover, to discriminate againstthe less profitable (i.e., more costly) patientswithin a specific DRG, two conditions would haveto hold. First, the physician would have to be ableto predict with reasonable accuracy the relativecostliness of different patients within the sameDRG at the time of admission; and second, thephysician would have to be induced not to ad-mit his or her more costly (and presumably sicker)patients. These conditions are simply unlikely tobe met frequently.

Of course, a hospital could simply choose notto participate in the DRG payment system by re-fusing all such patients. While this response is in-feasible in an all-payer system, it might be attrac-tive to some hospitals in a Medicare-only system.Total nonparticipation would be financially at-tractive to a hospital if the average DRG paymentlevel were to lie below the additional (marginal)costs of serving patients in any DRG, but it isunlikely in the foreseeable future that the paymentlevel, which is calculated on the basis of fullyallocated average operating costs, will be less thanmarginal costs for all DRGs in most hospitals. Ahospital could decide that the losses in some DRGsoutweigh the surplus available in others, but withMedicare accounting for about 30 percent of hos-pitals’ revenues, this situation would also be rare.Thus, the probability that many hospitals willrefuse to serve any DRG patients at all is low.

Constraints on Financial Incentives

Whether the financial incentives to reduce LOSand the cost per case and to increase admissionswill lead hospitals to overadmit patients and

28

underprovide services is an empirical question.The potential is real, but the possibility of adverseeffects on access and quality of care under DRGpayment is moderated by several built-in con-straints whose strength is unknown at present.

First, the physician, not the hospital admin-istrator, makes the decision to admit and dis-charge patients and order procedures. The physi-cian’s income often is dependent on hospitaliza-tion, as in the case of surgical admissions. Physi-cian visits to hospitalized patients may be morelucrative relative to their time requirements thanare office visits (28). Perhaps most important, thephysician’s professional and ethical standards pro-tect the patient from the withholding of neededcare. And, in a DRG payment system not cover-ing all payers, the physician would still be likelyto engage in a uniform style of practice for allpatients.

It is often asserted that defensive medicine—practices that are employed directly in responseto fears of malpractice lawsuits—would limit thewillingness of physicians and hospitals to engagein practices that threaten the outcome of care (91).The strength of the influence of malpractice onphysician behavior is arguable. There are no directobjective data on how much defensive medicineis practiced today or how much it costs. Physi-cians have claimed in some surveys that they per-form more tests than they otherwise would (67,85); in other surveys that they perform fewer tests(29) due to malpractice lawsuits.

Hospitals themselves are subject to malpracticesuits, which have risen dramatically since the first

Table 1 .—Malpractice

lawsuit was decided against a hospital in 1961(60,66). Approximately 75 to 80 percent of allmalpractice claims arise from medical care pro-vided in hospitals (60). An Institute of Medicine(IOM) study found in 1978 that a relatively smallnumber of institutions had formal programs formanaging such risks (60), but their frequency andimportance is growing (71). Even if objective es-timates were available on the extent of defensivemedicine and risk management under present con-ditions, it would be dangerous to generalize theseresults to a DRG payment system, where the fi-nancial incentives conflict with the incentives topractice defensive medicine. Thus, at this time,one can only conjecture about the potentialstrength of defensive medicine.

To the extent that it does function as a deter-rent to the underprovision of services, defensivemedicine may be less effective in protecting theelderly or disabled. There is a general consensusamong experts that these patients are less “liti-gious, ” in that they are less likely to sue physi-cians if they are harmed. A commonly cited rea-son for this is the fact that malpractice lawyerswork on a contingency basis and rarely acceptcases in which the claimant would not receive alarge compensation award. Most elderly and dis-abled persons would be awarded less money thanyounger patients, because part of the compensa-tion award is based on lost wages (60), and theelderly and disabled generally have lower incomepotential. Table 1 shows that the elderly receivedless money in closed malpractice cases in 1978 re-gardless of the severity of the injury suffered (61).

Claims Paid, 1978

Age of injured person

29

The incentive to increase admissions could con-ceivably be limited by the reluctance of patientsto be hospitalized for marginal indications or besubjected to the “revolving door. ” The Medicarebeneficiary is currently responsible for a deducti-ble of $304 upon hospitalization (18). It might beargued that this financial disincentive to hos-pitalization would moderate the incentive to ad-mit Medicare patients. Yet the deductible is notlikely to act as an effective deterrent to hospitaladmission. First, approximately 65 percent ofMedicare beneficiaries have private supplemen-tary insurance (“Medigap” coverage) which oftenpays for part or all of the deductible. Second, theelderly patient is unlikely to question “doctor’sorders” in a decision involving hospitalization.Third, the deductible will not adequately discour-age readmission because the beneficiary is liablefor the deductible upon a readmission only if itoccurs more than 60 days later than the previousepisode of hospitalization.1

In summary, natural limits do exist on the in-clination or ability of hospitals to overadmit,discharge too rapidly, and underprovide services.Yet, the magnitude of these constraints is un-known, and the protection of the elderly in par-ticular may be relatively weak. Programs to mon-itor hospital performance may be necessary toidentify behavior that is ultimately costly or harm-ful resulting from the economic incentives inherentin DRG payment.

Key Features of DRG Payment SystemsThat Affect Hospital Incentives

The effects of any DRG payment system on de-cisions in hospitals and, hence, on medical tech-nology use are influenced by five critical elementsof program design:

1. the proportion of the hospital’s case loadcovered by DRG payment,

Z. the treatment of costs as pass-throughs,3. the methods of DRG rate construction,4. the methods of updating DRG rates, and5. the level of risk and reward built into the

payment system.

The Proportion of the Hospital’s Case LoadCovered by DRG Payment

Every case excluded or exempted from the DRGpayment system will weaken its incentives. Ex-clusion of major payer categories from the system,for example, will limit its leverage on hospitals. *

Under the new Medicare law, about 32 percentof the revenues of non-Federal short-term hos-pitals will be subject to DRG payment (27), ex-cept in the few States with DRG payment systemscovering other payers as well. DRG paymentcould become even less important if States devel-op alternative prospective payment systems, asthe law allows them to do. Yet, a payment systemwith control over about one-third of hospitals’revenues is not inconsequential. To the extent thatit does force changes in hospitals’ behavior anddoes not merely shift costs to other payers, Med-icare’s DRG payment system will influence the useof medical technologies by all kinds of patients.Many changes in physicians’ practice patterns orhospitals’ purchasing decisions will probably beapplied broadly across all patients. And, if ahospital decides to eliminate a service to discour-age unprofitable DRGs, the service would be un-available to all patients.

The leverage of a DRG payment system can bereduced by exclusions built into the system itself.For example, recognition of “outliers,” cases withunusually high or low resource use, may reducethe strength of DRG incentives. Approximately20 percent of all cases in New Jersey’s all-payersystem, comprising 35 percent of hospitals’ costs,fall into the system’s outlier category (15). Andthe State’s criteria for declaring a case an outlierhave become more generous over time (97).

The treatment of outliers complicates hospitals’incentives. Exclusion of low-cost cases from DRGpayment is an important strategy for discourag-ing potentially unnecessary hospitalizations, par-ticularly for surgery that could be performed onan outpatient basis. Otherwise, in DRGs contain-ing both simple and complicated procedures, thehospital will have an incentive to admit the sim-ple surgeries as inpatients. At the high end of the

‘Social Security Act, sees, 1861(a) and 1813(a).

*Failure to cover all payers has important implications for equi-ty among payers, but that topic is beyond the scope of this mem-orandum.

30

cost distribution, there may be incentives to in-crease LOS in order to qualify a patient as anoutlier. The strength of these incentives dependson the location of the cutoff points-whether theyinvolve only a few or many patients—and thepayment method for outliers. New Jersey’s DRGsystem, which pays both high and low outlierson the basis of controlled charges, provides anincentive to manipulate LOS in high-cost patients.The Medicare system, which as currently legis-lated has no low-cost outliers, may encouragepotentially unnecessary admissions.

DRG payment systems could conceivably ex-clude certain types of cases or DRG categories onthe rationale that these categories or services needto be treated in a special way. Patients treated inburn care centers or psychiatric services, for ex-ample, could be excluded from DRG payment onthe grounds that these patients present uniquemedical and social problems. However justifiedsuch exclusions are, they would neverthelessweaken the impact of DRG payment and, depend-ing on how they are paid, could encourage ad-missions in these categories.

Treatment of Costs as Pass-Throughs

“Pass-throughs” are elements of hospital costthat are not controlled by the per-case paymentsystem. Cost-based reimbursement, as a whole,can be interpreted as a payment method in whichall cost categories are passed through. Per-casepayment systems that directly link a hospital’s per-case rates in a given year to its own previousyear’s costs are only minor departures from pass-through payment. Effective removal of pass-throughs requires a break in the link between thehospital’s own current costs and its future rate ofpayment.

Individual cost categories are treated as pass-throughs to varying degrees under different DRGpayment systems. During its first 3 years of opera-tion, Medicare will treat capital costs (deprecia-tion and interest payments) as complete pass-throughs: the hospital will be reimbursed forwhatever capital costs are incurred. New Jerseyhas established a capital facilities allowance forbuildings and fixed plant and equipment that isdesigned to meet the hospital’s need for cash to

pay off existing debt and to fund the downpay-ment for replacement or additions approved bythe health planning agency. For major movableequipment, such as beds and laboratory instru-ments, the State allows a depreciation rate thatis adjusted for inflation in replacement costs.Thus, except to the extent that the State’s healthplanning agency limits bed expansion or the ac-quisition of equipment, the New Jersey systempasses through capital costs.

Like New Jersey, Maryland has specific capitalallowances, but in the case of major movableequipment, the hospital’s asset value is calculatedin a base year and adjusted in subsequent yearswith inflation factors. The allowance for movableequipment is unaffected by the hospital’s subse-quent capital expenditure decisions except forspecial cases in which the ratesetting commissionmay make exceptions (42).

Other common pass-through categories underper-case payment are the costs of medical educa-tion (i.e., stipends of interns and residents, andteaching faculty costs), malpractice premiums,and utility expenses. Treatment of one or morecategories of cost as pass-throughs under DRGpayment renders these inputs to patient care freeto the hospital at the same time that the effectiveprice of all other inputs has been increased becauseof their inclusion in a per-case prospective system(46). In the absence of other effective controls,this change in the relative price of inputs giveshospitals an incentive to expand pass-throughinputs.

The Medicare law also excludes from per-casepayment an important product of hospitals: out-patient services. These services will continue tobe reimbursed on a retrospective cost basis (andthe patient is responsible for 20 percent co-insurance). Consequently, hospitals have a strongincentive to increase outpatient service volumesas a way of shifting fixed and overhead costs frominpatient to outpatient categories. Ancillary de-partments, such as radiology, clinical laboratory,physical therapy, and occupational therapy, willbe encouraged to compete for business with in-dependent providers of these services. New hos-pital-based home health services, which alsoescape the DRG system for now, are strongly en-

31

couraged both for their contribution to profitabili-ty and their prospects for reducing inpatient LOS(7,39,51,54).

Methods of DRG Rate Construction

The methods used to construct the relativeweights or prices of each DRG can affect hospitals’incentives. The important issue is how the ratioof cost to price varies among the patients servedby the hospital. This ratio of cost to price shouldbe constant across all patients; if not, incentiveswill exist to manipulate case load (i.e., to en-courage low-cost or discourage high-cost admis-sions). Though it is virtually infeasible to devisea per-case payment system that does not havesome variation in the ratio of cost to price, themethod of rate construction determines how greatthe variation is and which patients are paid ac-cording to relatively high and low rates.

There are two sources of variation across pa-tients in the ratio of cost to price: within-DRGvariation and across-DRG variation. Within-DRGvariation stems from the inherent heterogeneityof patients’ resource needs in a particular DRG.Any per-case payment method that establishes asingle price (or weight) for all patients in a case-mix category will result in some within-group dif-ferences in the ratio of cost to price. This cannotbe avoided, but the extent of the problem maydepend on the case-mix classification system. Therelative performance of DRGs and other case-mixclassification systems with respect to within-groupvariation has been discussed in chapter 2. Themethod of DRG price or weight construction doesnot alter this kind of variation. Policies regardingthe handling of “outlier” cases, discussed earlier,are more germane to this issue.

Conversely, across-DRG variation is deter-mined largely by the method of construction ofrelative DRG prices or weights. In theory, relativeDRG prices should reflect the relative costs of ef-ficient and clinically optimal patient care acrossDRGs (5,69). This would encourage hospitals tospecialize in those services that they can provideefficiently and to search for ways to further reducecosts. In practice: however, efficient care is dif-ficult to identify and even harder to measure, andat present, all DRG rates are constructed from em-

pirical estimates of DRG costs. The DRG case-mix index under TEFRA and the DRG prices ofthe new Medicare system are estimated from theaverage operating costs in a national sample ofMedicare hospital claims. Maryland uses the hos-pital’s own average revenue per case in a fixedbase year to develop relative DRG weights specificto the hospital. New Jersey combines averagestatewide costs with the hospital’s own averagecost of treating each DRG to arrive at a hospital-specific price. *

None of these methods assures that the relativeweights reflect efficient relative costs. Supposethat patients in one DRG are treated relatively ef-ficiently and uniformly throughout all hospitalswhile those in another are subject to a great dealof inefficient care. By Medicare’s average costcalculation, the inefficient DRG would be assigneda higher rate than it should be relative to the effi-ciently produced DRG. It is important to recog-nize both the reality of this problem and the op-portunity for mitigating it over time. As hospitalsrespond to the incentives of DRG payment, in-creases in their efficiency can be expected. Overtime, as DRG relative prices are recalibratedusing hospitals’ updated cost data, the disparitiesin cost-to-price ratios should diminish. Withoutrecalibration, whatever disparities in cost-to-priceratios existed at the beginning will remain.

The method used to allocate hospitals’ costs toparticular DRGs presents a more enduring prob-lem for relative prices. The Medicare method relieson hospitals’ charges to reflect average costs,where the DRG weight construction method isbased on the hospitals’ charges for services. Thesecharges are deflated by hospital-and-department-specific cost-to-charge ratios calculated from theMedicare cost reports. While this deflator reducessome of the distortions created by interdepartmen-tal subsidies, there remains a residual cross-sub-sidy of procedures and cases within departments.Cohen has claimed that this method compressesthe relative weight scale by underestimating thetrue cost of complex cases and overestimating thetrue cost of simple cases (9). Routine care ischarged at a flat rate per day, regardless of case

*See app. C for details.

98-823 0 - 83 - 6

-..——

32

severity, and some ancillary services, such as theuse of operating rooms, are billed on the basis oftime, not on the basis of resources needed to con-duct more complex procedures. Though the ex-tent of the bias is unknown, it implies that thecharge-based cost weights are likely to penalizethe more complex DRGs.

New Jersey intends to improve on this methodby directly observing the use of nursing time bypatients in various DRGs in selected hospitals (22).Direct observation of resource use is costly andhas some methodological problems (94), but theresults of these studies should provide valuableinformation on the magnitude of this problem.

Methods of Updating Relative DRG Prices

As the cost of efficient care in each DRGchanges over time, so, too, should the relativeDRG price. If it were reasonable to expect thatcosts would increase or decrease uniformly acrossall DRGs, then the only issue would be whetherthe average payment level is sufficiently high tocover the costs of efficient operation. But, uniformcost increases are highly unlikely: From year toyear, some DRGs will experience cost-saving tech-nological innovations; others will experience cost-raising ones. The relative prices of inputs (per-sonnel, supplies, energy, etc. ) also change, withconsequences for relative DRG costs. In the ab-sence of any changes in DRG prices, the ratio ofDRG price to efficient cost would show increas-ing divergence across DRG categories. As theseratios diverge, certain DRGs will become moreprofitable, others less so, and hospitals will havegreater incentives to engage in patient selectionstrategies. Therefore, the mechanisms employedto update, or recalibrate, relative DRG prices in-fluence the Iongrun incentives of the system. Re-calibration must depend on information if it is toavoid being completely arbitrary; thus, these up-dating mechanisms must include specification ofthe data and information systems available to sup-port them.

There are three basic approaches to recalibrat-ing relative DRG payment rates: empirical costestimation techniques, central policy decision ad-justments, and provider appeals.

All DRG pricing systems have originally beenestablished with empirical estimates of the relativecost of various DRGs. Periodic reestimation ofrelative costs based on updated data merely re-peats the process at reasonable intervals. NewJersey employs a ratesetting method that, at leastin theory, annually reestimates relative DRGcosts. The Medicare law calls for changes in DRGrelative rates at least every 4 years, but themethods to be used to recalibrate DRGs are un-specified. The law establishes an independentpanel of experts—the Prospective Payment As-sessment Commission—to recommend changes inrelative prices to the Secretary of the Departmentof Health and Human Services (DHHS) who willauthorize the changes. Presumably, the methodsused by the Commission will include reestimationof DRG costs.

Central policy adjustments in DRG rates occurwhen those in charge of ratesetting determine thatcertain changes in relative prices are justified totake account of new technology or changes in clin-ical practice. The Prospective Payment Assess-ment Commission is specifically charged withmaking recommendations about such adjust-ments. The Commission, and DHHS, will there-fore require an information base that exceeds thatneeded for empirical cost estimation. Data on thecost and clinical effectiveness of new technologywill also have to be collected and synthesized.

Provider-initiated appeals or petitions forchanges in relative rates represent the third avenuefor relative DRG rate adjustments. Like policy ad-justments, provider appeals can be used to adjustrates for changing technology, but this approachallows more flexibility in responding to the needsof particular hospitals. The burden of producingdata to justify changes in DRG prices rests to agreater degree on the appealing institution. NewJersey has instituted a DRG appeals mechanismto specifically account for changing technology.The new Medicare system prohibits appeals ofrates per se, but it does permit hospitals to ap-peal for additional payments for “outlier” caseswhose estimated per-case costs are extraordinarilyhigh. The effective price of DRGs containing newtechnologies (e.g., organ transplants), may bealtered through this process.

33

Risk and Reward

The degree to which the hospital is able to gen-erate surplus revenues and appropriate them toits own use will influence the strength of incen-tives to provide technologies more efficiently andcan also affect the hospital’s access to sources ofcapital. The ability to generate surplus dependson both the average level of payment and the rulesgoverning hospitals’ ability to keep surplus andliability for deficits. One program may emphasizethe risk side, putting hospitals entirely at risk forlosses without allowing them to keep surplus,while another may offer both substantial risks andrewards.

Traditional cost-based reimbursement is essen-tially a “no risk/no reward” system. DRG pay-ment systems vary widely in this regard. Hospitalsin New Jersey and Maryland can keep any sur-pluses attained from cutting costs per case andmust bear the full burden of cost increases. How-ever, both systems limit the revenue gains or lossesattributable to changes in admissions to their es-timated marginal costs or savings. In New Jersey,the potential for continued surplus-building insubsequent years is reduced somewhat by periodicrecalibration of DRG prices reflecting changes incosts. In Maryland, however, the benefits of costreductions (and the penalties for cost increases)are maintained in subsequent years, because DRGweights are not updated. Under the temporaryprovisions of TEFRA, the hospital reaps littlereward for keeping its per-case costs low (a max-imum of 5 percent of its per-case rate) but bearsthe full penalty of exceeding the per-case limit.Under the new Medicare law, the hospital bearsthe full burden of a loss and reaps the full rewardsof a surplus, regardless of their source. The hos-pital keeps the full portion of any surplus due toincreases in admissions. Thus, under the Medicarelaw, hospitals will have strong incentives both toreduce costs and increase profitable admissions.

Technology-Specific Effects ofDRG Payment

How do the general incentives of DRG paymenttranslate into specific effects on the use of medicaltechnologies? The previous sections demonstrateboth the complexity of the underlying incentives

and the impact of program design on their direc-tion and strength. DRG payment will not havea uniform effect on medical technologies and insome instances technologies will be subject to con-flicting incentives. From the discussions above itcan

●

●

●

●

b; concluded that:

Overall, the number and intensity of an-cillary procedures provided to inpatients canbe expected to decrease, but the use of pro-cedures that can be shown to lower the costper case will increase.The settings of technology use are likely tobe influenced by DRG payment, but the in-centives work in conflicting directions andare sensitive to the key features of programdesign. In the absence of an outlier policy forlow-cost patients, DRG payment encouragesinpatient admissions for simple procedures.On the other hand, the exclusion of outpa-tient costs gives hospitals an incentive to of-fer outpatient procedures. It remains to beseen which incentive will dominate for whichprocedures. DRG payment will encouragethe movement of technologies, particularlythose for posthospital care, into the homeand other nonhospital sites of care.DRG payment is likely to influence the spe-cialization of services, but the magnitude anddirection of these effects is unknown. The in-centives to reduce costs encourage concen-tration of capital-intensive technologies infewer institutions. Conversely, the increas-ing competition among hospitals for physi-cians and patients will create incentives forthe widespread acquisition of some technolo-gies.A change in technology product mix is like-ly to result from downward pressure on theprice and quantity of supplies and, if capitalis included in the DRG rate, capital equip-ment. Greater product standardization canbe expected as more expensive models andprocedures are eased out of the marketthrough competition.

Implications for Utilization Review andQuality Assurance

Per-case payment introduces much needed in-centives for cost control in hospitals, but it also

—

34

has potential negative implications for quality ofcare, access to care, and systemwide costs. Theincentives in DRG payment for hospitals to poten-tially manipulate case load, overadmit patients,discharge patients too early, and underprovide an-cillary technologies argue for safeguards in theform of quality and utilization review.

Review functions under Medicare have alwayshad two partially conflicting objectives: qualityassurance and cost containment. Under DRG pay-ment, these dual objectives remain. Utilizationreview will be necessary both to avoid costly in-creases in admissions and readmissions, and quali-ty audits will be necessary to protect inpatientsfrom the underprovision of technologies and fromtoo early discharge.

Both types of review overlap because of thetradeoff between quality and cost that becomesmore explicit with per-case payment. For exam-ple, physicians may become more selective in theirordering of diagnostic tests. Some tests may addto the cost per case but give better patient out-comes. Other tests can be avoided with little con-sequence for outcomes. Review processes that rec-ognize the balance between cost and quality be-come critical under DRG payment.

Historically, the responsibility for quality as-surance and utilization review has been shared byhospitals, intermediaries, and Professional Stand-ards Review Organizations (PSROs). Hospitalshave been required to have programs of qualityassurance and utilization review as conditions ofparticipation in the Medicare program2 as well asfor accreditation by the Joint Commission on Ac-creditation of Hospitals. Before the PSRO law wasimplemented, Medicare fiscal intermediaries wererequired to perform independent utilization re-views and thereafter remained the reviewers oflast resort in areas without active PSROs (68).Congress instituted the PSRO program in the 1972Social Security Act amendments (Public Law 92-

’42 CFR 405J.

603), establishing independent physician revieworganizations with the dual objectives of qualityassurance and cost containment.

In 1982, Congress replaced the PSRO programwith utilization and quality control peer revieworganizations (PROS) (Public Law 97-248). PROSwill be physician organizations whose perform-ance will be evaluated by the degree to which theymeet objectives for quality assurance and costcontainment specified in 2-year contracts withDHHS. Under the new Medicare DRG paymentsystem, hospitals must enter into agreements withPROS for review of the quality of care and theappropriateness of admissions and readmissions.

The integration of cost-containment and qual-ity-assurance objectives in a single physician-runindependent review organization such as a PROis both necessary and troublesome. Because theinherent tradeoff between cost and quality isbound up in every review decision, it would beimpossible to separate the two. Yet, it is difficultfor those responsible for conducting review andfor those funding such efforts to maintain a bal-ance between the two objectives. The history ofPSROs is instructive. Although the original in-tent of Congress was that PSROs were to bothcontain costs and assure quality, Federal evalua-tions of the program focused largely on the cost-containment objectives (86,87,88,91). The difficul-ty of specifying and measuring criteria for quali-ty of care added to the relative obscurity of thisobjective. The critical question to Federal policy-makers was whether PSROs were cost saving tothe Medicare program—i. e., did they reduce in-patient hospital utilization sufficiently to coverthe program costs? On the other hand, at the locallevel, PSROs emphasized quality assurance (25,59).

Whether PROS can strike an appropriate bal-ance between the cost containment and qualityassurance objectives remains to be seen. It is im-portant that at the Federal level the real need forquality assurance presented by DRG payment berecognized.

Chapter 4

Effects of DRG Payment onTechnological Change

in Medicine

Chapter 4

Effects of DRG Payment onTechnological Change in Medicine

INTRODUCTIONPerhaps even more important than how Diag-

nosis Related Group (DRG) payment affects theuse of presently available medical technologies ishow DRG payment will affect technologicalchange in medicine—the adoption of new tech-nologies and discarding of old ones. As a socie-ty, we value technological progress, the “introduc-tion to practice of new and more useful ways ofserving human purposes” (73). Technologicalchange should act as a filtering process, continual-ly winnowing out the less useful in favor of moreuseful technologies. To what extent can DRG pay-ment be expected to improve or hinder this proc-ess? Since DRG payment will influence hospitals’incentives to adopt new medical technologies, itmay therefore ultimately alter the rate and direc-tion of technological change in medicine. Thischapter will examine the implications of DRGpayment for technological change.

Conversely, technological change occurring dueto other forces will inevitably require adjustmentsin a DRG payment system itself, if the system isto continue to function effectively. These ad-justments range from changes in the relative ratesof payment across DRGs to changes in the defini-tions of DRGs themselves. Thus, a DRG paymentsystem must encompass procedures for timely ad-justment to new knowledge and technologicalconditions as they arise. This chapter will explorethe reasons for this conclusion and will also ex-amine the kinds of information and proceduresthat will be needed to monitor the performanceof a DRG payment system over time.

EVIDENCE OF EFFECT ON TECHNOLOGICAL CHANGEEmpirical evidence on the effect of DRG pay-

ment on the adoption of new technologies is un-available, but studies of other kinds of prospec-tive payment systems suggest that hospitals canand do respond to changes in financial incentivesin their decisions regarding the adoption of newtechnologies. Three studies of the impact of hos-pital prospective payment programs on the adop-tion of new services and technologies provideevidence of the response of hospitals to changesin financial incentives.

Joskow (41) analyzed the effect of ratesettingprograms on the availability of computed tomog-raphy (CT) scanning in hospitals. The number ofCT scanners in a State in 1980 was found to benegatively related to the number of years thatratesetting had been in effect in the State. Hospitalratesetting also led to a shift in the location of CTscanners to physicians’ offices.

Cromwell and Kanak (11) recently analyzed theimpact of 15 State ratesetting programs on theavailability of 13 different services in the hospitalbetween 1969 and 1978. Table 2 summarizes theresults. New York had the most consistently nega-tive effects on the availability of all types of serv-ices. New Jersey’s DRG ratesetting program alsoappeared to generally reduce the availability ofcomplex services, with the exception of elec-troencephalogram (EEG) services. Other States’programs showed no consistent impact on serviceadoption. It is interesting to note that the serviceupon which ratesetting had the highest and mostwidespread negative effect is social work, a labor-intensive, not equipment-intensive, hospitalservice.

In still another study of hospital payment andtechnology diffusion, Wagner, et al. (95), in-vestigated the impact of prospective payment in

37

38

no te*o

I

I

I

I

I

I

I

I

I

I

I

I

I I

I

I

I

I

I

I

I

I

I

I

o

I

I

n.

II

I II

I

I

I

I I

I I0I

I

I l l

u n~

I 01 I I II :I

I I I

I

%o0

I

I

I

I

I I II

I

I

I

I

I

I

I

I

I

II

I

I

I

I

I

vmo0 c

o

0coo0

I

(au)

o

I

I

I

I

I

I

I

I

I

I II I

I II III I I I

aI no0

mo0 I

I

o8

1 0

. .

oI

II

39

three States (New York, Maryland, * and Indiana)on the adoption of five new pieces of capitalequipment: electronic fetal monitoring, gastroen-doscopy, volumetric infusion pumps, automatedbacterial susceptibility testing, and computerizedenergy management systems. The first three tech-nologies probably raise the daily cost of care,although their effect on the average cost per caseis unknown. The latter two are investments incost-saving equipment. The New York ratesettingsystem was found consistently to lead to adop-tion of fewer units of the cost-raising technologiesand to increase the probability of large hospitals’

*The test of the Maryland system did not discriminate betweenhospitals paid by the case and those on the State’s conventional rate-setting system.

adopting the cost-saving equipment. However,the prospective payment programs in Marylandand Indiana showed no such consistent effects onhospitals’ adoption behavior.

Together these studies imply that prospectivepayment does affect the adoption of new technol-ogy in predictable ways, but that much dependson the strength and design of the program. NewYork’s system, the oldest and most restrictiveratesetting program, has clearly altered the extentof availability of new technology. Since its incep-tion in 1970, New York’s system has put a heavyemphasis on financial risk to the hospital whileat the same time offering little financial reward.Other systems may be too new, too small, or tooweak to show longrun consequences.

CLASSIFICATION OF NEW HOSPITAL TECHNOLOGIES

DRG payment incentives do not affect the in-troduction of new technologies uniformly. Thisis partly because of variation in the cost implica-tions of medical technologies. But it is also becausemedical technologies may be adopted (or notadopted) according to their clinical benefits (orrisks). These clinical implications naturally varywidely. However, for the purpose of analysis, itis necessary to classify new hospital technologiesaccording to criteria that highlight the likely ef-fect of DRG payment on their diffusion into thepractice of medicine. The most basic distinctionis due to the per-case nature of DRG payment.Thus, hospital medical technologies can be viewedas those whose adoption decreases the hospital’stotal cost per case—cost-saving technology-andthose whose adoption increases the hospital’s totalcost per case—cost-raising technology.

Cost-saving technology would include a newtechnology that provides a particular service lessexpensively, or one that reduces the required num-ber of services or length of stay (LOS) sufficient-ly to justify the investment. A cost-saving tech-nology in one hospital may be cost-raising inanother. The expected volume of use is often animportant factor in determining whether a tech-nology will be cost-saving. For example, capitalequipment that replaces other procedures may be

cost-saving in a large hospital but cost-raising ina small one. Had CT head scanning been intro-duced in an era of DRG payment, it may wellhave been justified on the basis of per-case costreductions in hospitals with large neurology ortrauma services (40,84). Yet its introduction intohospitals with lower neurological case loadswould probably have required justification onother grounds, such as improvements in patientaccess or outcome.

Many new procedures and therapies introducedin the past have raised the cost of hospital stays,even in high-volume institutions. Presumably,these cost-raising technologies improve patients’health outcomes or reduce patients’ medical careexpenditures outside of the hospital. Of course,some new cost-raising technologies have neitherimproved patient outcomes nor saved systemwidemedical care costs. Gastric freezing, a technologyof the mid-1960’s is a good example (21). Somehave claimed that intermittent positive pressurebreathing (IPPB), a respiratory therapy technol-ogy, provides another example (74).

Hospital technologies can be further classifiedaccording to their effects on capital and operatingcosts per case. Table 3 describes four kinds oftechnological innovation categorized by their ex-

40

Table 3.—Four Types of Technological Innovations and Effectson Capital Cost, Operating Cost, and Total Cost

Direction of effect on:

Capital cost Operating cost Total costType of innovation per case per case per case

1. Cost-raising, quality-enhancingnew technology . . . . . . . . . . . . . . . . . . . . . . . . . . + + +

Il. Operating cost-saving innovationsA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + — +B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + —

Ill. Capital cost-saving innovations—

A. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - + +B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - +

IV. Service/procedure disadoption . . . . . . . . . . . . . -——

SOURCE: Office of Technology Assessment.

pected effects on capital, operating, and total costsper case. Type technology, which raises all com-ponents of hospital cost per case, represents theclassic cost-raising, quality-enhancing technologythat would be introduced for its presumed patientbenefits. Intensive care monitoring is an exampleof such a technology. Type II represents a broadrange of capital investments that would save op-erating costs during the patient’s stay. Automa-tion technologies fall into this category, but so,too, do new diagnostic or therapeutic proceduresthat reduce LOS or intensity of care. These “op-erating cost-saving” technologies may or may notlower the total per-case cost of care, dependingon the relationship between capital and operating

costs. Type III technologies involve new, simplerapproaches to care, in which expendable labor,materials, or supplies are substituted for capitalequipment. The abandonment of a capital cost-saving technology falls into this category, as doesa new labor-intensive test that replaces less effec-tive but automated laboratory procedures. Final-ly, Type IVrepresents the disadoption of a serv-ice or procedure resulting from new knowledgeof its lack of clinical efficacy or safety. The rapiddecline in recent years in the use of IPPB therapyfollowing publication of evidence on its lack ofefficacy in many clinical settings is a goodexample.

GENERAL INCENTIVES FOR TECHNOLOGICAL CHANGEIN DRG PAYMENT

Although there is no empirical evidence, it ispossible to describe the incentives regarding tech-nological change under DRG payment. Many ob-servers have pointed out that per-case paymentsystems, in which future levels of payment arelargely independent of the hospital’s own his-torical costs, create incentives for hospitals toadopt cost-saving technologies. Yet because tech-nologies are neither cost-saving nor cost-raisingindependent of the context in which they are used,all hospitals cannot be expected to adopt the sametechnologies. The introduction of new capital-intensive cost-saving technologies in a DRG pay-ment environment is likely to speed up the proc-

ess of specialization as small hospitals becomeunable to reap the cost-saving benefits of someinvestments. Some technologies that depend onhigh volume to be cost-saving could be providedto smaller hospitals on a contract basis by largehospitals or independent laboratories. The feasi-bility of such arrangements would vary with thespecific uses of a technology and the geographicaland competitive environment in which the hos-pitals operate.

The introduction of new “cost-raising” technol-ogies is discouraged, but not eliminated, underDRG payment compared to cost-based payment.

—

41

Under cost-based payment, the additional hospitalcosts of new technologies are fully covered; hos-pitals therefore have no financial incentives notto adopt such innovations. Under DRG payment,the adoption of new cost-raising technology is notmet with an automatic increase in revenues tocover the additional cost. New technology willhave to compete with alternative uses of funds,such as employee wage and benefit increases, ad-ditional nursing staff, etc. New technology maybeat an additional disadvantage relative to otheruses of funds because of the relative uncertaintyabout its benefits in the early stages of diffusion(72). The implications are obvious: with limitedresources, hospitals will need to assess new tech-nologies more closely and ration resources morecarefully.

Nevertheless, the introduction of promisingnew technologies, even those that are cost-raising

to the hospital, will be attractive to hospitals asthey compete for physician loyalties and, ulti-mately, the admissions they represent. For exam-ple, despite its high capital and operating cost,nuclear magnetic resonance imaging, a new med-ical imaging technology still in the investigationalphase, may be highly desirable to hospitals thatseek to protect their admissions base from en-croachment by other hospitals. The importanceof this incentive as a constraining force to theprevious incentive is unknown. Thus, thoughDRG payment, in general, does not imply thattechnological change will aproach a standstill, itsdirections are likely to be altered, and the adop-tion of technologies that are cost-raising to thehospitals is likely to decline by an unknown quan-tity.

KEY FEATURES OF PROGRAM DESIGN THAT AFFECTTECHNOLOGICAL CHANGE

As with its impact on technology use, the im-pact of DRG payment on technological changewill be influenced by the key features of programdesign. In particular, the methods of providingrewards for cost reductions, treating capital costs,and updating DRG prices have important implica-tions for technological change.

Risk and Reward

The ability of hospitals to generate and keepsurplus sufficient to fund investments in new tech-nology depends to a large extent on the averagepayment level of the payment system as a whole.A DRG payment system can be either restrictiveor generous. A restrictive payment policy will en-courage rapid adoption of cost-saving technolo-gies but will discourage investments in cost-raisingtechnologies. A more generous payment level mayreduce the pressure for the adoption of cost-savingtechnology but will also provide surplus for cost-raising investments.

Treatment of Capital Costs

The exclusion of capital costs in the DRG pay-ment rates will change hospitals’ incentives to in-troduce certain kinds of new technologies. * Table4 describes how the incentives provided by DRGpayment are influenced by the capital paymentmechanism. Capital payment methods do notalter the direction of such incentives as long asthe effect of a new technology on total cost percase is in the same direction as its effect onoperating costs. For example, DRG payment pro-vides a disincentive to adopt most cost-raising,quality-enhancing (Type I) technologies regardlessof the way in which capital is handled. Capitalcost pass-throughs weaken the disincentive notto adopt this kind of technology, but they do not

*Exclusion of costs from DRG payment rates is referred to as treat-ing those costs as “pass-throughs. ” Under some DRG payment sys-tems, pass-throughs have been subjected to different kinds of con-trols. (See app. C for a description of New Jersey’s and Maryland’ssystems. )

. — —

42

Table 4.—impact of Technological Innovation on Per-Case Costs

Direction of effect on: Incentives for adoption

Capital cost Operating cost Total cost With capital without capitalType of innovation per case per case per case in rate in rate

1. Cost-raising, quality-enhancingnew technology . . . . . . . . . . . . . . . . . . . . . . . . . + + + 1 1

Il. Operating cost-saving innovationsA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + — + 1 !B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . + — — f f

Ill. Capital cost-saving innovationsA. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – + + 1 1B. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . – + — I 1

IV. Service/procedure disadoption . . . . . . . . . . . . – — t 1SOURCE: Office of Technology Assessment.

remove it. New technologies with high capitalcosts, but only small increases in operating costs,would be affected less by DRG payment with acapital cost pass-through than by DRG paymentwith capital built into the rate. Similarly, capitaldisinvestments (Type IV), occasioned perhaps bythe introduction of a simple procedure to replacea capital-intensive one or simply by the abandon-ment of an ineffective technology, are encouragedby DRG payment regardless of the way in whichcapital costs are treated.

Operating cost-saving (Type II) and capitalcost-saving (Type III) technologies can lead tosituations where the incentives for hospitals toadopt may actually be reversed by the policy re-garding payment of capital. Of particular concernis the incentive under a pass-through to adopt ex-pensive capital equipment that reduces operatingcosts but raises total cost per case. Like regulatedutilities, where allowed profits are computed asa percentage of the amount of equity capital, hos-pitals can be expected over time to become toocapital-intensive (3).

The Social Security Amendments of 1983 (Pub-lic Law 98-21) provide for capital to be treatedas a pass-through under the new DRG paymentsystem. Capital technologies will continue to bepaid retrospectively on the basis of reasonablecosts. However, the inadequacy of this methodwas recognized by Congress in the new law. Thelaw requires the Department of Health and Hu-man Services (DHHS) to submit a report to Con-gress by October 20, 1984 on payment for hospi-tal capital. The intention of the law is that at theend of 3 years, Congress will have legislated a new

capital policy that will deviate from present cost-based reimbursement. Although the issues incapital formation policy are complex and beyondthe scope of this technical memorandum, anypolicy that eliminates the pass-through will havedesirable effects on the consistency of hospitaltechnology acquisition decisions.

Methods of Updating DRG Pricesand Categories

The longrun viability of any DRG payment sys-tem depends on its ability to both adapt to andencourage appropriate technological change inmedicine. A payment system that is rigid in theface of medical progress will become unacceptableto providers, patients, and the public. Conse-quently, the methods and procedures used to ad-just the average payment level, relative DRGrates, and the DRG categories themselves are crit-ical to the longrun survival of the system.

The primary objective of a DRG price adjust-ment process is to maintain equality across DRGsin the ratios of price to the cost of efficient care.This objective implies that as new cost-savingtechnology becomes available for use in certainDRGs, the relative prices of these DRGs shouldbe adjusted downward to reflect the new efficien-cies. Alternatively, the development of new cost-raising technologies that improve patient out-comes enough to justify their use should be metwith increases in the prices of relevant DRGs. Ad-justing relative prices to reflect technologicalchange as it occurs provides appropriate incen-tives for efficiency and specialization in thedelivery of hospital care.

43

But to simply adjust prices in reaction to tech-nological change is insufficient. A second objec-tive of the price adjustment process is to encouragehospitals to adopt appropriate new technology.The previous sections concluded that while DRGpayment encourages the adoption of cost-savingtechnology, it reduces hospitals’ incentives toadopt new cost-raising technologies. Dependingon the general restrictiveness of the paymentsystem, the effects may be to discourage diffusionand perhaps even the development of costly buteffective new hospital technologies. The DRGprice adjustment process can be used to counteractthis incentive for deserving technologies.

There are two general methods of DRG priceadjustment: conditional adjustments and uncon-ditional adjustments. Conditional DRG rate ad-justments are those granted only to hospitals ac-tually adopting a new technology; unconditionalrate adjustments are those that apply to all hos-pitals (or to a class of hospitals) whether or notthey use the new technology. There are three pos-sible unconditional adjustment processes:

Periodic empirical reestimation of relativeDRG costs. —This method is statistical andreactive in nature; as technological changealters the costs of serving specific DRGs, thecalculated rates change. The process is alsogradual, because estimated relative weightsbased on average costs across both adoptersand nonadopters of a new technology wouldonly partially reflect the effects of the newtechnology on the efficient cost of acceptedcare. Only in the final stages of diffusionwhen the new technology is uniformly ap-plied across all hospitals would the estimatedrelative costs reflect the technology’s fulleffect.Application of a technology factor. —Thismethod employs an annual percentage in-crease in the average rate of payment for allDRGs to provide funds for the adoption ofcost-raising technology. * For example, thenew Medicare law requires that the annualincrease in the average payment level reflecttechnological change as well as general in-

*In theory, a negative technology factor could be used to reducefunds in anticipation of cost reductions, but it has not been appliedin this manner in any State system.

●

flation, but the amount of increase is notstatutorily specified. Inclusion of such a fac-tor in the annual rate increase gives hospitalsfunds to use for any purpose, including tech-nology adoption.Central policy decision to change relativeDRG rates. —-A designated State or nationalagency can make purposeful adjustments inrelative DRG rates to reflect a change in tech-nological conditions. The new Medicare lawspecifies that the Prospective Payment As-sessment Commission make recommenda-tions regarding periodic adjustments in rel-ative DRG rates to reflect changing technol-ogy. This Commission could recommend anincrease in the rate paid for a particular DRGrelative to other DRGs as a means of fund-ing the acquisition of a cost-raising technol-ogy. Of course, hospitals treating patients inthe DRG with the increased rate would befree to use the additional revenue for anypurpose.

Conditional adjustments in DRG prices can beaccomplished by two general mechanisms:

●

●

Provider-initiated appeal. —Individual hos-pitals contemplating adopting a new technol-ogy can request an adjustment of rates in spe-cific DRGs to fund its acquisition. Presuma-bly, hospitals would not request reductionsin rates due to cost-saving innovations. In-stead, this approach is potentially useful forcase-by-case exceptions to DRG rates to payfor new cost-raising technology. The newMedicare law specifically prohibits hospitalsfrom appealing DRG rates, but appeals areallowed for very high cost “outlier” patientson a case-by-case basis. The State of NewJersey has a DRG appeals mechanism that isintended to bring to the surface new technol-ogies in need of extra payment. To qualifyas a DRG appeal, the new technology mustbe shown to affect one or more DRGs ac-counting for at least 10 percent of the hos-pital’s costs or admissions and to affect oneor more hospitals other than the applicant(63).Creation of new DRGs. —New DRGs, dif-ferentiated from preexisting ones by the useof a specific technology, can be created as

——

44

a way of paying a hospital only if it actuallyadopts and uses the technology. The newDRG would be assigned a price reflecting thehigher or lower resource costs associated withthe use of the new technology. For example,liver transplantation might become its ownDRG, although the Medicare program doesnot yet pay for such procedures. When andif the new procedure is considered ready forpayment, a DRG price will be assigned thatcan be obtained by any hospital performingthe procedure. Those hospitals not perform-ing transplant operations will not receive theadditional revenue, because they will haveno patients in that DRG.

The first objective of DRG rate adjustment—the maintenance of equal ratios of price to the costof efficient care among DRGs—depends largelyon unconditional rate adjustments. Although peri-odic reestimation of DRG costs represents a grad-ual adjustment to new technology, it is a reason-able method for adapting to cost-saving techno-logical imovation. Because cost reductions wouldbe reflected only gradually in relative rates, hos-pitals adopting such innovations would reap apositive, albeit declining, return on their in-vestments over a long period. Hospitals laggingin adoption of cost-saving innovation would grad-ually be subjected to greater penalties. Thus, thereestimation process is a gentle adjustment methodwhich nevertheless embodies strong incentives forhospital efficiency.

Central policy decisions to force more rapid ad-justment of payment rates to new cost-savingtechnologies are also feasible and may be usefulfrom time to time. Yet, it maybe difficult earlyin the availability of a new technology to predicthow large a saving to expect and how such sav-ings are likely to vary among hospitals. Since theincentives to adopt new cost-saving technologyare already strong in DRG payment, radical re-ductions in specific DRG rates are likely to bemore disruptive than useful.

The second objective of a DRG adjustmentprocess-encouraging the adoption of appropriatetechnologies, particularly those that are cost-raising to the hospital-may use either conditionalor unconditional adjustment methods. But no one

method alone is completely satisfactory for thispurpose. Periodic reestimation of DRG costs isnot likely to be sufficient to induce hospitals toadopt desirable cost-increasing technologies, es-pecially very expensive ones. Early adopterswould bear the full extra costs of such a new tech-nology, but the updated weights based on averag-ing costs across both adopters and nonadopterswould not reflect the full increase in per-case costs.Hence, the initial stages of the diffusion processwould be underfinanced.

Inclusion of a general technology factor in theannual average rate of increase does provide fundsfor the adoption of new cost-raising technology.However, this kind of across-the-board increaserewards innovative and noninnovative hospitalsalike and may even cushion some hospitals againstthe need to become more efficient, because theyare free to spend the additional revenue howeverthey choose. Consequently, a technology factoris an inefficient approach to encouraging theadoption of cost-raising technology, since it islikely to fund less of such innovation than the costof the approach to third-party payem.

The same criticism can be made of central pol-icy decisions to increase the relative rates of cer-tain DRGs. The hospital would be free to use theextra payment from an adjusted DRG price forany purpose. Thus, centrally mandated rate ad-justments would give equal reward to adoptersand nonadopters.

Yet, conditional adjustment mechanisms, suchas creation of new DRGs or individual providerappeals, have their own shortcomings. Creationof new DRGs may appear on the surface to bean ideal approach, but it has serious deficienciesas the major tool. The prospect of DRG infla-tion—the gradual increase in the number of cat-egories-has severe implications for managementof data and information. If the experience withmedical procedure nomenclature is any guide, therate of increase in the number of categories canbe expected to be high (55). The first revision inDRG nomenclature increased the number of cat-egories by 22 percent. The additional data re-quired on hospitals’ claims to make fine distinc-tions in DRG assignments would certainly add anadministrative burden on hospitals and payers.

—.

45

Second, and more important, over time DRGpayment would come to look more and more likefee-for-service medicine, where the amount ofpayment is inextricably linked to the proceduresperformed. Financial incentives to perform prof-itable combinations of procedures could develop.In addition, substitution of a new cost-saving pro-cedure for a more expensive one would be dis-couraged if it were to bump a patient out of thehigher priced category into a lower priced one.Thus, the uncontrolled expansion of categoriescan create a more rigid, less cost-effective healthcare delivery system.

Reliance on case-by-case hospital appeals ofDRG rates is likely to be administratively costlyand cumbersome compared to other methods. Un-fortunately, there is little experience to date withNew Jersey’s DRG appeals process, and even ifthere were, it is unclear whether the experienceof a single State is germane to operation of a na-tional appeals process. Nevertheless, the NewJersey appeals process bears watching as a poten-

tial model for adoption in a more general DRGpayment system.

It is worth noting that any approach to up-dating relative DRGs to account for new cost-raising technology requires information on whichto base conclusions about the readiness of a newmedical technology for payment. In effect, tech-nology assessments are needed to support the deci-sion process. For example, should the relativerates of those DRGs using hyperalimentation beadjusted to reflect the additional costs of this tech-nology? Such questions would of necessity trans-cend case-by-case appeals processes. A mecha-nism and adequate resources for conducting in-tegrated and comprehensive assessments of suchquestions are important for supporting the DRGadjustment process. The Prospective PaymentAssessment Commission established by the Med-icare law is empowered to conduct or fund suchassessments, but other research resources may alsobe usefully employed in providing information tosupport such critical decisions.

COVERAGE POLICY, DRG PAYMENT, AND TECHNOLOGICAL CHANGE

Since its inception, the Medicare law has man-dated that a specific technology or service mustbe a “covered” benefit, i.e., a benefit eligible forpayment, in order for providers to be reimbursedfor its provision. Although Medicare specifiesbroad categories of covered benefits, specific tech-nologies, particularly new ones, require individualcoverage determinations. Such coverage decisionsare governed by section 1862 of the Social Securi-ty Act, which excludes payment for items andservices that are “not reasonable and necessary”for diagnosis, treatment, or improved services.“Reasonable and necessary” lacks precise defini-tion: the Medicare program applies the terms totechnologies that are generally accepted by themedical community as being safe and effective andare perceived to have moved beyond experimen-tal status to clinical application.

These coverage requirements have acted as abarrier to adopting new, untested technologies

and, in a passive way, have discouraged the aban-donment of outmoded technologies. Although theextent to which coverage policy has been an ef-fective “technology watchdog” varies among par-ticular technologies, it is generally accepted as in-fluencing the diffusion of technologies. Under thenew Medicare law, the provisions of section 1862remain. Since DRG payment will also affect therate and direction of technological change, therelation of the two, coverage policy and DRGpayment, has implications for medical technol-ogy in the Medicare program.

At this time, the interactions between Medicarecoverage policy and DRG payment as mandatedin Public Law 98-21 are limited to the hospital por-tion of inpatient services. As noted previously,outpatient services and physician services-pro-vided in or out of the hospital—are not includedin the DRG system. Instead, they are paid for asbefore the law’s enactment: outpatient services are

46

reimbursed on a reasonable cost basis, and physi-cian services are reimbursed on a reasonablecharge basis.

The ability of coverage decisions to assist in ap-propriate technological change depends in part onthe identification of new technologies and newuses of established technologies. If a technologythat is part of the hospital portion of inpatientservices (now paid by DRG payment) comes tothe Medicare program’s attention as “new,” adetermination of its safety and efficacy will stillbe required. Although DRGs are generally con-structed according to diagnosis (and not treat-ment), there are several ways to identify newtechnologies under the newly mandated DRGpayment system. A few still-to-be-covered tech-nologies, the most obvious being heart transplant,are specific DRGs. Program administrators alsowill have the potential to identify establishedsurgical procedures in reviewing claims for DRGpayment, but will not be able to identify newsurgical procedures by claims review.

Another opportunity for identifying new tech-nologies will occur when hospitals appeal theirpayment levels for “outlier” cases. Many of thesecases will be high-cost outliers precisely becausenew and costly technologies will have been em-ployed. If a new procedure is not covered, denialof an outlier claim is a likely prospect.

Finally, new technologies and new uses of es-tablished technologies will be recognized duringthe process of adjusting DRG rates. Indeed, up-dating DRG prices appears to offer the most sig-nificant opportunity of identifying such technol-ogies for coverage purposes. The decision to ad-just DRG rates can therefore be considered aquasi-coverage decision itself.

Under current Medicare coverage policy, oncea new technology is identified and brought to theattention of program administrators, the HealthCare Financing Administration (HCFA) arrangesfor the Public Health Service to conduct an assess-ment to ascertain the safety and efficacy of thetechnology. HCFA subsequently determines thecoverage status of the technology based, in part,on the results of the technology assessment.

For the DRG payment system, changes in DRGrelative weights or prices will be made, in part,to reflect technological change. Because thisprocess* must include identification of new tech-nologies, it is reasonable that some of the tech-niques, including technology assessments, used toadjust DRG rates will be similar to those used forsupporting coverage decisions. For example, theProspective Payment Assessment Commission hasbeen given broad powers to assess medical tech-nology and the appropriateness of medical prac-tice patterns in performing this task. It is thisCommission that must make recommendations tothe Secretary of DHHS concerning the appropri-ate payment rate for hospital services accordingto its findings.

Thus, both the coverage process and the proc-ess of adjusting DRG rates share a similar “ap-proval for payment” function. They differ in thata coverage determination focuses on specific tech-nologies while adjusting DRG payment rates fo-cuses on the larger entity of a diagnostic group,which includes particular technologies. Moreover,the DRG rate adjustment process must includeissues of cost as an integral issue, while thecoverage process at present does not consider costissues. Nonetheless, the technology assessmentsperformed for both processes no doubt will besimilar. The potential for duplication is not to beignored.

Whether technologies will be subject to a dou-ble review of safety and efficacy for payment pur-poses will depend on the approach chosen to up-date DRG prices. Irrespective of approach, it isreasonable to assume that hospitals’ adoption ofcost-raising technologies will be made evident toHCFA for DRG payment and for coverage policy.However, some approaches to updating DRGrates, i.e., provider appeals, would not surfacecost-saving technologies. Because specific tech-nologies will not be identified on the DRG hospitalclaims form, the adoption by a hospital of a new,uncovered, cost-saving technology would notbecome known to HCFA through hospital claims

*As noted in the previous section, the specific mechanisms of theprocess have not yet been determined.

47

review. However, as in the past, HCFA will relyon physician claims and other sources to provideinformation to stimulate the initiation of a tech-nology assessment solely for coverage purposes.

Many new technologies, especially those thatare cost-raising, could conceivably be faced withthe prospect of a double barrier to introduction—an assessment for coverage purposes and an as-

sessment for DRG rate adjustment. Although theevaluations may differ for the two purposes, theprocesses appear to be sufficiently similar to war-rant coordinated Government effort. The unin-tended confluence of the two administrative proc-esses needs to be examined in order to conserveFederal resources and to promote the diffusion ofappropriate technologies.

Chapter 5

Selected Implementation Issues:The Hospital Perspective

Chapter 5

Selected Implementation Issues:The Hospital Perspective*

The preceding chapters have examined poten-tial hospital behavior toward technology use andadoption when faced with a prospective per-casepayment system based on Diagnosis RelatedGroups (DRGs). As noted earlier, potential be-havior toward technology is predicted to vary asspecific features of a system vary. In addition tothe general and specific incentives provided by thesystem, a number of issues concerning implemen-tation arise that should be noted by policymakers.These issues assume even greater importance inview of the recently legislated Medicare paymentsystem. Thus, this chapter will briefly examineissues involving the implementation of a DRG re-porting system to support per-case payment. Itwill not focus on implementation issues from theperspectives of the Health Care Financing Ad-ministration (HCFA) of the Department of Healthand Human Services or the intermediaries whowill also be involved in the operation of the newsystem.

PATlENT DATA ISSUES

Classification and Coding Errors

Because assignment of patients to DRGs re-quires data from the patients’ discharge abstracts,the accuracy and timeliness of these data havecome under question. Several studies have beenundertaken under the auspices of the Institute ofMedicine (IOM) of the National Academy of Sci-ences to determine just how accurate and timelythe data are, and how various procedures mightbe employed to ensure data reliability (57,58,59).

In each study, a sample of patient records wasreabstracted by a trained field team of RegisteredRecord Administrators and compared with theoriginal data compiled by either the abstract serv-ices (for the first study), HCFA (for the Medicaredata in the second study), or the National Center

● This chapter incorporates work by Diane E. Hamilton, RalphE. Berry, and Nancy L. Kelly of Policy Analysis, Inc.

Implementation issues encompass a variety ofproblems with varying cost implications. Two im-portant aspects of implementation of the newMedicare DRG payment system will be describedin this chapter: data and coding issues and hospitaladministrative issues. Data issues exist because ofthe reliance of the classification process on datasummaries (discharge abstracts) of the patients’hospital medical records. These issues includeclassification and coding errors, and DRG “gam-ing” or “creep. ” Administrative issues exist be-cause the administrative burden to hospitals ob-viously increases when any new type of paymentsystem is introduced, particularly one as complexas a DRG payment system is likely to be. For ex-ample, hospitals will need to institute proceduresto improve the quality of information, includinga reduction in missing data in medical charts.

for Health Statistics (NCHS) (for the third study).For diagnosis and procedures, two types of datadiscrepancies were possible. Ordering discrepan-cies would reflect problems in determining whichof several diagnoses or procedures should be re-garded as principal. Coding discrepancies wouldreflect errors in assigning a diagnosis or procedurecode number.

Findings from these studies indicated that hos-pital data on admission date, discharge date, andsex were highly reliable; however, this was notthe case when diagnosis and procedure data wereexamined. For all diagnoses combined, whencodes were compared (up to four digits), Medicaredata were reliable in only 59.5 to 64.1 percent ofthe cases. In the study of abstract service data,the comparable figures had been 66.8 to 77.5 per-cent, and in the NCHS study, 63.4 to 86.0 per-cent. Further, Medicare data concerning the pres-

51

52

ence of additional diagnoses were reliable in 74.5percent of the cases, and the reliability level forMedicare principal procedures was 78.9 percent.Finally, in the abstract service and NCHS studies,the field team concluded that in 4.6 percent of thecases, the correct diagnosis code was a matter ofjudgment. This was also true for 1.7 percent ofall procedures in the Medicare study.

It should be noted, however, that the data usedin IOM studies were for 1974 and 1977. A secondand even more important consideration is that thestudies were based on detailed coding of diagnosesand procedures; therefore, discrepancy rates didnot reflect error rates that might occur when caseswere aggregated into DRGs. In fact, in a studyof coding error at the DRG level, reliability in-creased to 76.7 percent (69). A third problem withextrapolating from these studies is that the dis-charge abstracts studied were not produced forpayment purposes. When payment depends onthe accuracy and timeliness of discharge abstracts,their importance increases and data reliabilityshould improve. Monitoring by peer review or-ganizations (PROS) in the new system should givehospitals added incentive to improve their datacollection and coding procedures.

The dependence of payment on coded diag-noses and procedures in a DRG payment systemraises the possibility of deliberate changes incoding conventions. Several authors (4,77,91)have noted that the ability to maximize paymentby changing diagnosis codes could be a seriousproblem. “DRG creep” was defined by Simborgas a deliberate and systematic shift in a hospital’sreported case mix in order to improve payment(77). As described in chapter 2, the primary basisfor subdivision of cases into discrete DRGs is theprincipal diagnosis. Using the original DRG clas-sification system, Simborg pointed out that bychanging the sequence of discharge diagnoses forpatients with more than one diagnosis, a higherpriced DRG can result. If done systematically,perhaps using sophisticated computer programs,a more costly case mix would result.

It should be noted that the potential for “codingcreep” exists with all available case-mix measuresand is potentially even more problematic withthose requiring subjective severity determination.To some extent, the new DRGs limit the discre-

tion permitted in the assignment of DRGs, reduc-ing but not eliminating the “upcoding” possibilitiesthat the original DRG system offered. Under themodified DRG system, only significant, predeter-mined secondary diagnoses (complications andcomorbidities) or age can lead to a case being in-cluded in a higher cost DRG; i.e., sequence nolonger matters. In addition, a surgical procedurehierarchy is now used to assign patients who hadsurgery to DRG categories. Where there are multi-ple medical and surgical conditions, the one in-volving the major surgical procedure becomes theprincipal one. Thus, surgery takes precedence.

Data processing sophistication should increasewithin the hospitals in response to the new Med-icare payment system. For example, it would paya hospital to use its computer to screen for un-complicated cases or certain DRGs that the med-ical records department should review for poten-tial undercoding (77). DRG creep, or deliberateovercoming, can be controlled in two ways. First,PROS or other review organizations can screencertain DRGs for overcoming. This function wasspecifically assigned to PROS by the new Medicarelaw. Second, the potential gains from DRG creepwould diminish if DRG prices are regularly rees-timated. New prices or weights would reflect thenew distribution of patients among DRGs and thenew average costs per DRG. Over time, reestima-tion of weights would cause the more profitableDRGs to become less profitable, and the less prof-itable ones more profitable. Thus, one could ex-pect a gradual decline in the potential for “gam-ing” via DRG creep with periodic reestimation ofDRG prices.

Clearly, these improvements can be expectedfrom the hospital industry as administrators, med-ical records persomel, and particularly physiciansbecome more aware of the reimbursement impli-cations of inaccurate data. Obviously, improve-ments of this type have potentially significantresource implications for the hospital industry,as well as third party payment agencies. Thesecosts should be considered when the potential im-pact of a DRG reimbursement system is assessed.New Jersey has conducted educational programsfor medical records personnel, physicians, nurses,and hospital administrators. Data accuracy hasimproved subsequently (22,76), though someDRG creep may exist in New Jersey (14).

53

HOSPITAL ADMINISTRATIVE ISSUES

The foregoing discussion of problems in dataaccuracy and timeliness in the use of DRGs is in-dicative of the need for improvements in the pro-cedures used for data abstraction and coding. Re-searchers who have examined some of the prob-lems have suggested several areas in which im-provements should be made.

First, because a great deal of the error is in-troduced at the hospital level, programs to im-prove the quality of the information should beinstituted. These might include additional train-ing for persons abstracting information from themedical record, routinization of hospital pro-cedures so that activities of billing personnel couldbe limited to information transfer (rather than in-terpretation of the medical record data), and in-structional programs for physicians in classifyingdiagnoses, determining principal diagnosis, andcompleting the medical record* (12). Again, NewJersey has implemented these suggestions and hasfound them to be successful (22,48,49).

The medical record should also be completedin a timely fashion in order to bill third partiesas soon after discharge as possible. In this case,physicians must be encouraged to complete themedical record as soon after discharge as possi-ble. Also, for some hospitals, additional medicalrecords personnel may be necessary.

A third suggestion for improving data qualityis to establish direct, timely links among errordetection, feedback, and training (10). (In fact,the New Jersey system has instituted many editingand educational initiatives throughout that State. )It is suggested that this error detection should in-clude, as a supplement to data checks by the com-puter, a regular program for independently reab-stracting samples of records. New Jersey Profes-sional Standards Review Organizations have donethis to monitor DRG assignments of patients toDRGs (14). As stated earlier, however, these types

● Demlo and her colleagues (12) have also suggested that if thepractice of determining principal diagnosis by referring to the first-listed item on the face sheet of the medical record continues, themedical record format itself might be revised so that the conditionsare recorded in order of priority with the principal diagnosis listedfirst.

of improvements are not without cost or time im-plications, and there is some evidence to suggestthat these improvements may increase the averagecost of preparing a bill under a DRG-based pay-ment system.

Some preliminary results of the effect of DRGpayment on hospital behavior and performanceare available from the New Jersey DRG paymentexperience (98). To assess the effect of the ex-perimental DRG-based payment system on hos-pital organization and procedures, comparisonswere made between matched samples of partici-pating and nonparticipating hospitals. Based onthat comparison, the following conclusions werereached:

1.

2.

3.

A

the importance of the medical records de-partments has increased dramatically in par-ticipating hospitals. This was considered tobe the result of the required expansion of thedepartments’ functions and personnel, aswell as the need for better trained personnel;the medical staff in the participating hospitalshas become more directly involved in hos-pital operations; andthe quantity and type of information col-lected in DRG hospitals has expanded, al-lowing for the development of more sophis-ticated management information systems.

second portion of the New Jersey analysisexamined the quality and timeliness of data beforeand after the institution of the new payment sys-tem and found that while data accuracy im-proved, the length of time needed to produce thedata increased. Although the abstract face sheetincompletion rate decreased, the amount of timerequired by the medical records and patient bill-ing departments increased. The author stated thatas hospitals become more experienced with thesystem, it is likely that the time needed to proc-ess data will decrease (98).

The details of the methods of the New Jerseyevaluation are not yet publicly available, so thatinterpretation of the results must be preliminary.Nevertheless, there is some indication that imposi-tion of a DRG-based reimbursement scheme re-quires additional administrative resources. The

—

54

magnitude of these additional resources and the in current hospital procedures. It is hoped by pro-implications for Medicare payment cannot nec- ponents that the increased administrative burdenessarily be inferred from this preliminary analysis. would be offset by the cost savings attributableSome hospitals could be expected to incur larger to the imposition of a flat-rate, prospective pay-cost increases than others as a result of differences ment system.

Chapter 6

Findings and Conclusions

Chapter 6

Findings and Conclusions

DRG EVALUATION ISSUES

Diagnosis Related Groups (DRGs) were devel-oped as a method for characterizing hospitals’ casemix using data commonly available on hospitaldischarge abstracts. Although the motivation be-hind their original development was unrelated totheir use in hospital payment systems, they wereready for implementation at a time when a con-sensus was developing that hospitals should bepaid for their outputs (treated cases) rather thanfor their inputs (services or days of care). As theonly system of case-mix classification currentlypractical for use with per-case payment, it is un-derstandable that DRGs have been selected for usein the Medicare program.

DRG PAYMENT ISSUES

Although there exists little empirical evidenceabout the effect of DRG payment on the adop-tion and use of medical technology, potentialproblems with various aspects of program designcan be identified.

First, just as cost-based reimbursement hascreated inappropriate incentives regarding the useand adoption of medical technology, so too doesDRG payment, but the incentives are new and dif-ferent. Whereas overuse of inpatient services andtoo lengthy hospitalization were problems undercost-based reimbursement, underprovision ofservices and inappropriate admission and dis-charges may be a problem under DRG payment.These incentives will require programs of quali-ty assurance and utilization review designed spe-cifically to deal with them.

Second the incentives affecting the use of med-ical technology depend on several important as-pects of system design. In particular, the way inwhich capital costs are treated will affect incen-tives to use and adopt medical technologies. Indrafting the Medicare law, Congress recognizedthat treating capital costs as a pass-through itemis not an optimal longrun approach. Hospitals are

Nevertheless, it is important for those whowould use DRGs to recognize that in their pres-ent state of refinement (i.e., the 467 DRG clas-sification) they have been inadequately evaluatedfor their validity as an indicator of patient re-source needs and for their impact on medical tech-nology in per-case payment. In light of the budgetcrisis facing the part A Medicare trust funds inthe upcoming years, to move to DRG paymentis reasonable. But given how little anyone knowsabout what to expect from DRG payment, it iscritical that its implementation be carefullymonitored, particularly with respect to its effecton the use and adoption of medical technology.

likely to become too capital-intensive over timeas a result. The diversity in hospitals’ ages, debtstructures, and future needs for expansion orclosure all argue for hospital-specific determina-tions of the capital payment levels. These issuesare inevitably intertwined with planning for healthfacilities and are therefore most amenable to treat-ment on a State level.

Third, DRGs must be updated to both reflectand induce desirable technological change if thesystem is to remain responsive to the needs of allpatients. Periodic reestimation of relative DRGrates to reflect changes in the costs of variousDRGs is essential to a workable program. Rees-timation guards against growing divergence in theratios of DRG cost to DRG price, and it alsocounteracts the potentially deleterious effects ofDRG creep. How frequently such reestimationshould occur is a debatable issue. The new Medi-care law mandates recalibration at least every 4years, but this interval may be too long. Morefrequent, perhaps annual, reestimation has disad-vantages in increased administrative burden onprogram administrators and reporting require-ments on hospitals. But these administrative costs

57

58

might be offset by the enhanced ability of the Fed-eral Government to capture cost savings as theyoccur and by the strengthened incentives to adoptcost-saving innovations more quickly. Annual re-estimation would also be more effective in con-trolling the longrun incentives for DRG creep thanwould infrequent reestimation. Whatever its in-terval, reestimating relative DRG costs implies theneed for a continuing source of cost and chargedata to support the process. Plans for alteringMedicare billing forms and cost reporting re-quirements should proceed with these require-ments in mind.

Methods for updating DRG rates that are con-ditional upon technology adoption may be im-portant to stimulate desirable but cost-raising in-novations. The adjustment process should allowfor differentiation in rates between adopters andnonadopters of new medical technologies whosediffusion needs to be stimulated. Creation of newDRGs and provider appeals represent the onlyviable conditional adjustment methods, and eachof these has shortcomings. In particular, heavyreliance on new DRGs runs the risk in the longrun of creating a fee-for-service system in thehospital, the precise opposite of what DRG pay-ment is intended to do. Provider appeals conjure

up visions of administrative bureaucracy anddelays which detract from the otherwise attrac-tive simplicity of DRG payment. Yet, appeals maybean important vehicle for encouraging new tech-nologies. New Jersey’s DRG appeals mechanismshould provide some insight into its usefulness inthis regard.

The DRG adjustment process requires support-ing mechanisms for identifying and assessing newhospital cost-raising technologies. Judgmentsabout the readiness of new technologies for pay-ment under one or more DRGs need to be sup-ported by evidence about their effectiveness, risks,and costs. While the Medicare law established aProspective Payment Assessment Commissionwhose purpose, among others, is to collect suchinformation, adequate research resources arenecessary to support the process.

The reliance of the DRG classification systemon accurate and timely data collection and codingwill necessitate improvement of hospital’s medicalrecords procedures and performance. Educationalprograms for physicians, nurses, hospital ad-ministrators, and medical records personnelshould be initiated. Monitoring of informationquality both within hospitals and by the mandatedpeer review organizations will be necessary.

IMPLICATIONS FOR RESEARCH AND EVALUATION

Although a conclusion of this memorandum isthat DRGs are ready for use in per-case paymentand that they are currently superior for this pur-pose to any other measure, the importance of agood case-mix measure in making per-case pay-ment a viable longrun payment strategy impliesthat research on alternative measures must con-tinue. It is too early to consider DRGs the basisfor all future changes in case-mix measurement.The new Medicare law requires the Departmentof Health and Human Services (DHHS) to studythe appropriateness of modifying DRGs to ac-count for illness severity or other factors. Thisstudy should be enlightening, but by itself it isnot enough. DHHS has an excellent record forsupport of the development of DRGs and other

case-mix measurement techniques. Continued ag-gressive Federal support for development andrefinement of promising alternatives is requiredto reap real improvements in case-mix measure-ment techniques.

So little is known about the magnitude of theeffects of DRG payment on the utilization of serv-ices and technologies in the health care system thatsystematic study of these effects is needed. Studiesneed to be designed now to evaluate the effect ofDRG payment on rates of admissions, lengths ofstay, and the use of ancillary, outpatient, andnonhospital care. The new Medicare law man-dates a study of its impact on hospital admissions.Such a study could be part of a larger investiga-tion of the law’s effects.

59

Since the effects of DRG payment on access toand quality of care are unknown at present, thesefactors should be closely monitored as the Medi-care program is implemented. Patterns of servicespecialization are likely to change, and while theseresults may have benefits, they may leave pocketsof inadequate access in some areas. The hallmarkof the Medicare program has been the great in-crease in access of the elderly to mainstreammedical care. The effect of DRG payment on ac-cess, particularly through hospitals’ decisions toopen and close services, merits close scrutiny inthe coming years.

The importance of pass-throughs in altering theincentives of hospitals argues for careful study ofways to expand the scope of DRG payment. The

Medicare law mandates separate studies of twoimportant elements that are currently pass-through items: capital and teaching costs.

Finally, the remaining questions about the im-pact of specific elements of program design inaltering general hospital incentives provided byper-case payment make study of State-level alter-native prospective payment systems an attractiveprospect. Although States need not adopt per-casepayment, the law encourages them to design sys-tems that they can be reasonably confident willdo at least as well as the Medicare system in con-taining hospital costs. Evaluations of such sys-tems, as mandated in the new law, will provideimportant information to support future improve-ments in payment system design.

IMPLICATIONS FOR PAYMENT SYSTEM ORGANIZATION

The relatively untested introduction of a na-tional per-case payment system using DRGs as thecase-mix measure is a bold step toward improv-ing the efficiency of hospital care. But the systemas designed in the new Medicare law is imperfect,and the details of program administration are stillto be worked out. DRGs are the best case-mix in-dicators currently ready for use in a payment sys-tem, but there are other measures under develop-ment with equal or perhaps even greater promise.Moreover, there are many potential useful ap-proaches to prospective payment and even to per-case payment. It is therefore fortunate that thenew Medicare law does not discourage individualStates from establishing alternative prospectivepayment systems. These alternative systems willallow for experimentation with different paymentsystem configurations, including the use of othercase-mix measures as they become more refined.

By statute, alternative prospective payment sys-tems must cover a high proportion of the State’s

inpatient admissions. The inclusion of payers inaddition to Medicare in prospective payment willstrengthen its incentives. Furthermore, there aremany components of per-case payment that ap-pear to be suited to decentralized administration.For example, utilization review, provider-initiatedappeals, and decisions regarding payment for cap-ital costs are more amenable to decentralized ad-ministrative structures.

When these administrative issues are consideredin conjunction with the potential for sharing ad-ministrative costs with other payers, considerableattention should be given to the possibility thatthe future of DRG payment rests on the degreeto which the States join with Medicare to deviseall-payer systems. The incentives analyzed in pre-vious sections would all be strengthened under all-payer systems.

Appendixes

.

Appendix A.— Health Program Advisory Committeeand Acknowledgments

HEALTH PROGRAM ADVISORY COMMITTEE

Sidney S. Lee, Chair,Michael Reese Hospital and Medical Center

Stuart I-I. Altman*Florence Heller SchoolBrandeis University

Carroll L. EstesDepartment of Social and Behavioral SciencesSchool of NursingUniversity of California, San Francisco

Rashi FeinDepartment of Social Medicine and Health PolicyHarvard Medical School

Melvin A. GlasserHealth Security Action CouncilCommittee for National Health Insurance

Patricia KingGeorgetown Law Center

Joyce C. LashofSchool of Public HealthUniversity of California, Berkeley

Margaret MahoneyThe Commonwealth Fund

Frederick MostellerDepartment of Health Policy and ManagementSchool of Public HealthHarvard University

Mitchell Rabkin*Beth Israel HospitalBoston, Mass.

Dorothy P. RiceDepartment of Social and Behavioral SciencesSchool of NursingUniversity of California, San Francisco

Richard K. RiegelmanGeorge Washington UniversitySchool of Medicine

Walter L. RobbMedical Systems OperationsGeneral ElectricMilwaukee, Wis.

Frederick C. RobbinsInstitute of Medicine

Rosemary StevensDepartment of History and Sociology of ScienceUniversity of Pennsylvania

● Until April 1983.

63

64

ACKNOWLEDGMENTS

The development of this technical memorandum was greatly aided by the advice and review of a numberof people in addition to the Advisory Panel for the study on Medical Technology and Costs of the MedicareProgram and the Health Program Advisory Committee. OTA staff would like to express their appreciation especial-ly to the following individuals:

Laura Allendorf, American Society ofInternal Medicine

Gerard Anderson, Department of Health andHuman Services

Robert Berenson, Washington, D.C.Patricia Booth, Health Care Financing

Administration, Department of Health andHuman Services

Alan Cohen, Johns Hopkins Medical InstitutionsHarold Cohen, Maryland Health Service Cost

Review CommissionBarbara Cooper, Health Care Financing

Administration, Department of Health andHuman Services

Richard Crout, National Institutes of HealthHelen Darling, Government Research Corp.W. Palmer Dearing, Blue Cross/Blue Shield

AssociationAllen Dobson, Health Care Financing

Administration, Department of Health andHuman Services

Dennis Duffy, Suburban Medical ReviewAssociation, Kenilworth, N.J.

Albert Esposito, Health Care FinancingAdministration, Department of Health andHuman Services

Robert Evans, University of British ColumbiaRobert Fetter, Yale UniversityJoanne E. Finley, Maryland State Department

of HealthFaith K. Goldschmidt, New Jersey State

Department of HealthJoseph Gonnella, Jefferson Medical College,

Philadelphia, Pa.Lyla Hemandez, American Medical Peer Review

AssociationSusan Horn, Johns Hopkins Medical InstitutionsMark Hombrook, Kaiser-Permanente Medical Care

ProgramMichael J. Kalison, Manger, Kalison, Murphy &

McBride, Morristown, N.J.Jack Keane, Maryland Health Service Cost Review

Commission

. . .

Irvin Kues, The Johns Hopkins HospitalLeo Lichtig, Blue Cross of Northeastern New YorkBryan Luce, Health Care Financing

Administration, Department of Health andHuman Services

John E. Marshall, National Center for HealthServices Research, Department of Health andHuman Services

Katherine Means, Health Care FinancingAdministration, Department of Health andHuman Services

Mark Ozer, Washington, D.C.Dale Patterson, Department of Health and

Welfare, IdahoSteven Pelovitz, Health Care Financing

Administration, Department of Health andHuman Services

Julian Pettengill, Health Care FinancingAdministration, Department of Health andHuman Services

Ellen Pryga, American Hospital AssociationMarsha Rosenthal, Prudential Insurance Co.Leonard Saxe, Boston UniversityJo A. Surpin, Alpha Health ConsultantsJames Vertrees, Health Care Financing

Administration, Department of Health andHuman Services

Jeffrey Warren, New Jersey Rate SettingCommission

Jeffrey Wasserman, New Jersey Health Researchand Education Trust

Carolyn Watts, University of California,Berkeley

Karen Williams, Health Insurance Associationof America

Donald Woodbury, Georgia Department ofMedical Assistance

Donald Young, Health Care FinancingAdministration, Department of Health andHuman Services

Wanda Young, Blue Cross ofWestern Pennsylvania

Appendix B. —A Brief Review of theDevelopment of DRGs*

Overview and Historical Perspective

The development of Diagnosis Related Groups(DRGs) has been ongoing since the late 1960’s, and itis appropriate to view the concept as one that is con-tinuously evolving. To this point, the evolution ofDRGs has involved both conceptual refinements andtechnical improvements, spurred by the availability

of more and better quality input data and by feedbackfrom a wide variety of observers and users of DRGs.It is likely that the evolution will continue as relevantdata increase in availability and improve in quality andas the concept is subjected to more and more scrutiny.

The first version of DRGs to be widely disseminatedwas a set of 383 categories, described by their devel-opers in 1980 (19). Subsequently, in early 1982, a sec-ond and much revised set of 467 categories was issued(103). This revised set bore little resemblance to the“original” 383, as it was based on different definitionalprocedures and a different coding convention. Bothsets had several common objectives. Both were de-signed to identify patients with similar expectedresource use, measured by length of hospital stay. (Theadvantages and disadvantages of the length of staycriterion will be discussed subsequently. ) Both versionswere defined so as to be medically meaningful to physi-cians, the key decisionmakers within the hospital withrespect to patient care, though the operationalizationof this objective varied significantly between the two.Finally, both sets of DRGs were deliberately based ondata that are commonly available, and both setssought to be limited to “manageable” numbers ofgroups.

In general, the broad outlines for the constructionof both sets of DRGs were the same for each version.Actual patient stays in a sample of hospitals were theunits of analysis. Each patient’s principal diagnosis—i.e., the reason (after study) that the patient wasadmitted—was coded using a detailed coding systemthat allowed for many thousands of possibilities. Thefirst step, therefore, was to collapse the detaileddiagnosis codes into meaningful, but broad, sub-categories called “Major Diagnostic Categories”(MDCS). MDCs were then further subdivided, usinga combination of statistical analysis and medical judg-ment, according to other characteristics that accountedfor differences in resource use within the MDC.

The major differences, however, may appear to out-weigh the similarities. Significantly modified pro-

cedures were used to develop the 467 DRGs. These dif-ferences included the involvement of a far greaternumber of participants, many of them clinicians,which accompanied a shift in the fundamental orien-tation of the development process. Whereas the de-velopment of the 383 DRGs had involved both sta-tistical analysis and expert clinical judgment, thebalance between the two components was relativelymore even than it became in the revised method, inwhich the balance was shifted in favor of clinicaljudgment.

In addition, there were a number of differences inthe specific features of the development process. Thedifferences were so extensive that there is very littlecorrespondence between the two sets of DRGs. In theremainder of this appendix, the procedures used tocreate each of the two sets will be summarized and thesimilarities and differences among them will be ex-amined.

Development of the “Original”383 DRGs**

The original 383 DRGs were developed from datafor approximately 500,000 patients. Most of these werefrom New Jersey hospitals, though additional datawere also available from a large hospital in Connect-icut and for a sample of patients reviewed under theFederal Professional Standards Review Organizationprogram. Before the data were analyzed, cases thoughtto be misleading or unrepresentative were eliminatedfrom further consideration. These included deaths,miscodes, and patients with extremely long lengths ofstay (LOS). The reason for this exclusion was that theoverriding objective of the process was to describe a“typical” patient. Apparently, aberrant cases weredisregarded.

As a first step, clinicians classified the patient recordsinto 83 mutually exclusive and exhaustive MDCs.MDCs were based on both the etiology (or cause) ofthe disorder and the organ system involved. The 83MDCs thus contained a number of categories that wereapplicable to the same organ system. For example,MDCs relating to the respiratory system included ma-lignancies of the respiratory system, pneumonia, acuteupper respiratory infections and influenza, asthma,bronchitis, and other lung and pleural diseases.

● This appendix is based on a paper prepared for OTA by Nancy L. Kelly,Diane E. Hamilton, and Ralph E. Berry of Policy Analysis, Inc.

● ● Much of the substance of this section was derived from Fetter and col-leagues (19).

65

66

Analysis was then performed to determine if eachMDC should be further subdivided in order to reducethe variance in length of hospital stay. LOS, as notedearlier, was selected because it was viewed as a keyindicator of resource use and the best such indicatorfor which data were available. Justification for its usewas the close correspondence between LOS, case com-plexity, and cost that has previously been reported inthe literature (19). Particularly because of the per diembased rate structure of many of the existing reimburse-ment systems, LOS was considered to be a reasonablyaccurate and accepted measure of resource use. In ad-dition, the consistency with which this measure wasreported was considered a practical advantage. Moredirect measures of resource use, such as charges, were(and are) not only more difficult to obtain, but theyare more difficult to make comparable across areas,due to wage and price differences, and across hospitals,due to differing markups. Consequently, LOS becamethe focus of the statistical analysis and strongly in-fluenced the final form of the diagnostic categories.

Many have argued that while the number of daysa patient is hospitalized is one indicator of resourceconsumption, LOS by itself may not be an accurateindicator of total patient treatment costs (4,13). Thisbecomes even more evident when one considers dif-ferent treatment patterns, including the many differentancillary procedures possible, that may occur with dif-ferent patients in the same DRG.

The use of LOS as the primary measure of resourceconsumption also contributes to the lack of homoge-neity within the original DRGs formed. For instance,the old DRGs grouped together in a single categorylung cancer patients with a short diagnostic workup,a lengthy chemotherapy treatment, or a terminal ad-mission (4). Researchers have suggested that clinicaldata in addition to those already used in the originalDRG classification system are needed in order to con-struct groups that are more homogeneous from botha clinical and resource consumption standpoint (4,13).For instance, age, socioeconomic status, and type ofadmission have been suggested as important elementsin classifying patients into homogeneous groups (4).

Subdivision of MDCs resulted from an iterativeprocess, during which statistical output was reviewedby clinicians in order to determine which grouping al-ternatives best satisfied both medical and statistical cri-teria. The statistical analysis involved the assessmentof whether within-group variance in LOS was signif-icantly reduced when the patients were subdivided ac-cording to secondary diagnosis, primary and second-ary surgical procedures, and age. For example, if LOSamong all pneumonia patients was highly variable, but

it was much more similar within groups of elderly andnonelderly patients, then a decision would be madeto divide the MDC “Pneumonia” further, accordingto whether or not the patient was 65 or older. A nextstep might involve examining the improvement in LOShomogeneity (or, more technically, the “reduction invariance” in LOS) when elderly and nonelderly pa-tients were further subdivided according to the pres-ence or absence of a secondary diagnosis. The processwould continue until it was determined that furthersubdivision would not significantly reduce the vari-ance, would not be medically meaningful, would resultin too many groups, or would result in too few casescontained in a group.

The number of variables investigated was deliberate-ly limited to a small number to reduce the complexityof the analysis. However, they consisted of the vari-ables that the developers considered to be the key dis-criminators among the characteristics for which datacould be obtained from hospital records. Others, suchas sex, were tested but did not prove to be importantin explaining variations in LOS.

The process of defining DRGs was therefore a com-plicated one, involving both subjective and objectivetechniques. Although a consistent approach was main-tained within each MDC, few firm guidelines were em-ployed during the initial development phase. However,it appears that, during the formation of the originalDRGs, a slight edge was given to the statistical criteria.One result of this was that some MDCs, such as acutemyocardial infarction, were not subdivided at all andthat others, such as fractures, were subdivided into asmany as 13 DRGs. Also, this approach resulted in vari-ation across MDCs in the criterion (or criteria) for sub-division. For some, such as appendicitis, secondarydiagnosis was the criterion for subdivision; while forpneumonia, age, surgery, and secondary diagnosiswere all used. This difference was accepted insofar asit was seen to reflect variation in relevant patient char-acteristics.

The result was a set of 383 groups that simultaneous-ly satisfied the criteria established for distinctiveness,medical meaningfulness, and size. It is worth notingonce again that the 383 DRGs were based on input dataderived from a sample of patients that was mainly lim-ited to the northeast region. The result thus reflectedthe composition of cases in that sample, as well as themedical practice patterns employed in the hospitalsfrom which these patients were discharged. Also, judg-ments about the alternative grouping configurationswere made by a small group of clinicians whose viewsmay not have been representative of physicians na-tionally. The possibility that the initial set of DRGs

67

● Much of the substance of this section was derived from Yale University(103).

within the original DRGs. The 38 experts who par-ticipated in this study identified and ranked 16 cat-egories of problems that caused inappropriate assign-ments. Most important to them was that “some DRGscombine clinically similar patients, who neverthelessrequire different treatments” (99). They also identified37 of the original 383 DRGs as categories likely to con-tain patients whose clinical status was not appropriate-ly recognized. Physicians and hospital administratorsidentified different reasons for the problems, but theyspecified the same problematic DRGs (99).

Several of the problems identified by Williams andcolleagues concerned the medical meaningfulness ofDRG-based case-mix measures. Some who have usedthe original DRGs argue that they are not medicallymeaningful because patients with very different med-ical problems are grouped together (4). For instance,old DRG 301 groups together all patients whose prin-cipal diagnosis is “replacement of hip with prostheticdevice, biopsy of bone, and spinal fusion.” In addi-tion, DRGs fail to subdivide some broad diagnosticgroups. The original DRG 121, for instance, includesall patients with acute myocardial infarction.

As a result of such criticisms, significant changeswere made in the organization and orientation of thedevelopment process, the manner in which decisionswere made, and the nature of the input data. Spe-cifically, a more structured organization was used toadminister the classification process and to guide thedecisionmaking. A large number of participants fromthe medical profession, as well as other areas withinthe health industry (e.g., medical records) were in-volved. For this later phase, data were made availableby the Commission for Professional and Hospital Ac-tivities from the Professional Activity Study (PAS).This meant that a nationally representative sample ofpatients could be selected and analyzed, thus improv-ing the generalizability of the results.

DRG development procedures were substantially al-tered during this second phase. The major change wasin the basic orientation of the decisionmaking, in thatstrong emphasis was placed on the clinical, rather thanstatistical, validity of DRGs, The first manifestationof this change was in the redefinition of MDCs. Ratherthan using the 83 MDCs defined previously, the re-vised approach redefined a total of only 23 MDCs,most of which were confined to a single organ system.To return to the example used earlier, diseases of therespiratory system, which were represented by sixMDCs in the earlier methodology, were representedby a single MDC in the modified approach. Only afew MDCs (e.g., burns) remained the same. The pur-pose of this change was to bring MDCs into conform-ance with the organization of medical practice, inwhich, for the most part, specialties are defined around

68

the various organ systems. This also appeared to fa-cilitate the increased use of expert clinical judgmentin developing MDC subdivisions.

Additional changes were made in the process afterthe redefinition of MDCs. An important change in theanalysis was the retention of patients who died.Another was the development and use of more preciseguidelines for subdividing MDCs than had been usedpreviously. An outgrowth of this was more consisten-cy in the application of criteria for subdivision. Forexample, the guidelines required that the initial parti-tion (when possible) be based on the presence or ab-sence of a surgical procedure performed in an operat-ing room. The need for grouping patients with clinical-ly related diseases was continually stressed in theguidelines.

Several important modifications were made in thevariables used to subdivide MDCs, primarily to cap-ture severity of illness more precisely. One such mod-ification instructed the expert panels to rank ordersurgical procedures according to resource use intensi-ty and to assign patients with multiple procedures tothe procedure involving the greatest intensity. Thismeant that the type of surgical procedure became animportant consideration in the new DRG develop-ment, whereas in the original grouping procedure, onlythe presence or absence of surgery was taken into ac-count. In addition to considering the type of surgeryperformed, the nature of existing comorbidity (i.e.,coexisting conditions) and complications was explicitlyevaluated based on the specific ICD-9-CM codes con-tained in the patient’s discharge abstract. Again,“substantial” comorbidity and complications weredistinguished from those considered to be less signifi-cant. “Substantial” was defined to include conditionslikely to increase LOS by at least 1 day for at least75 percent of the cases. In many instances, a compositevariable indicating whether or not the patient was aged70 or more and/or had substantial comorbidity orcomplications proved to be an important determinantof resource use.

Finally, other variables in addition to diagnosis, pro-cedures, and age were taken into account when theexperts judged that the additional factors were rele-vant. For example, for the MDC “Pregnancy, Child-birth, and the Puerperium,” the initial division is madeaccording to whether or not the patient was “deliveredthis admission.” With respect to substance abuse, theinitial split is according to whether or not the patient“left [the hospital] against medical advice.” Death wasalso included as a possible criterion for subdivision.

While, as noted earlier, the modified process wasmuch more dependent on clinical judgment, statisticalanalysis again was used to aid decisionmaking. Reduc-

tion in variance for LOS was again examined for eachpartitioning variable considered, but the fact thatvariance was significantly reduced by a particularvariable did not guarantee that that variable wouldbe included in the modified DRG definition. Clinicalcoherence, above all, dictated which measures wouldbe used.

The modified approach resulted in the definition of467 DRGs, * which bear little resemblance to the orig-inal 383. To the extent that the original groups can be“mapped” into the revised ones, it is clear that whilesome of the original groups were further subdividedby the new process, others were collapsed into fewercategories. For example, the original DRG 121 was“acute myocardial infarction” (AMI), undifferentiated.In the new configuration, AMI patients are classifiedinto three DRGs:

● circulatory disorders with AMI and cardio-vascular complications, discharged alive;

● circulatory disorders with AMI, without cardio-vascular complications, discharged alive; and

. circulator disorders with AMI, expired.In an example of the opposite effect, bronchitis andasthma were divided into three bronchitis- and threeasthma-related DRGs under the original system. Themodified set includes only a total of three DRGs forbronchitis and asthma combined.

Comparison of Alternative Sets of DRGs

Table B-1 presents a summary of the fundamentalsimilarities and differences between the original 383DRGs and the modified set of 467. Most of the specificareas shown in the table have been discussed in theprevious section. This concluding section, therefore,will focus on the implications of the changes made.

Clearly, the major thrust of all of the methodologicalchanges was to improve the medical meaningfulnessof DRGs. To the extent that this was accomplished,it should result in the increased acceptability of theDRG scheme to physicians, who manage the medicalcare provided to hospital inpatients. As a consequence,the necessary interactions between clinical staff andhospital administrators should be improved. The de-velopment of a medically meaningful grouping schemehas always been a clear objective of those who origi-nated the concept of DRGs.

The enhancement of the “clinical coherence” ofDRGs was attempted in several important ways:

• by increasing the number of clinician participantsand the role of clinicians in DRG development;

● A 468th DRG has also been defined to account for patients who havereceived an operating room procedure unrelated to their MDC. This “outlier”category is not normally included in descriptions or evaluations of the system.

69

Table B.1.—Summary of Similarities and Differences Between the Original and Modified DRGs

Original DRGs Modified DRGs

Based on data for patients from New Jersey and Based on nationally representative sample of patientsConnecticut, and a sample of Medicare/Medicaid derived from PAS data (deaths retained)patients (deaths excluded)

Based on ICDA-8 (or HICD-2) diagnostic coding scheme Based on [CD-9-CM diagnostic coding schemeResult from subdivision of 83 broad subcategories of Result from subdivision of 23 broad subcategories of

diagnoses (MDCs) based on organ system and etiology diagnoses (MDCs) based on organ system onlySubdivision of MDCs based on statistical analysis and Subdivision of MDCs based on clinical judgment and

clinical judgment statistical analysisVariables used to subdivide MDCs include Variables used to subdivide MDCs include type of

presence/absence of secondary diagnoses and surgery surgery and comorbidity/complications, age, death, andas well as age other relevant criteria

Subdivisions not uniform across DRGs When possible, first subdivision based uponpresence/absence of operating room procedure;generally, tighter guidelines for subdivision wereapplied

End result: 383 mutually exclusive and exhaustive DRGs End result: 467 mutuallv exclusive and exhaustive DRGsSOURCE: Office of Technology Assessment.

●

●

●

by tying the initial subdivisions (MDCs) to majororgan systems, in conformance with the delinea-tion of medical specialties;by taking the specific nature of surgical pro-cedures and comorbidity/complications into ac-count in forming the groups; andby extending the number of characteristics usedto partition-MDCs when appropriate.

It would appear that significant strides have been madetowards the objective of medical meaningfulness.These would also seem to be important strides if physi-cian behavior is to be the target of management or costcontrol efforts.

It is unclear whether grouping according to clinicalsimilarities was attained at the expense of statisticalvalidity, and if so, whether it is important. The deem-

phasis of statistical analysis as a mechanism for deci-sionmaking in forming the groups implies that the newgroups may be less internally homogeneous and dis-tinct from each other than the original set. Admitted-ly, however, the basis for evaluation of within-grouphomogeneity and between-group heterogeneity—LOS—is an imperfect criterion for indicating resource con-sumption. Preliminary evidence does indicate, how-ever, that the groups achieve similar reductions invariance for charges as they do for LOS (103). Dataavailable for about 330,000 New Jersey discharges (in-cluding cost data) were analyzed, and the results in-dicated that the distribution of costs was similar to thatof LOS for most DRGs. In the few instances wherethere were significant differences, modifications weremade to the relevant DRGs.

Appendix C.— Examples of Per-Case andDRG Payment Systems

Diagnosis Related Groups (DRGs) have been usedin three State ratesetting systems, as well as in theMedicare reimbursement system under the Tax Equi-ty and Fiscal Responsibility Act of 1982 (TEFRA) andthe Social Security Amendments of 1983 (Public Law98-21). TEFRA was designed as a temporary responseto the problem of hospital cost-containment and spe-cifically called for the Secretary of the Department ofHealth and Human Services (DHHS) to develop a pro-posal for a permanent system of prospective paymentunder Medicare. A December 1982 proposal by theformer Secretary of DHHS, Richard Schweiker, calledfor the creation of a national DRG-specific paymentsystem for Medicare beneficiaries (91). In April 1983,a Medicare DRG payment system was enacted into lawwith features similar to those suggested in theSchweiker proposal. The new system will be phasedin over a 3-year period beginning in October 1983.

Theoretically, DRGs could be used in any hospitalpayment method, including retrospective cost-basedreimbursement, but their importance in paymentderives from their use as part of prospective per-casepayment systems. Per-case payment refers to any pro-spective hospital payment system with fixed rates ofpayment based on the hospital admission, not on thebundle of services or number of days of care provided.DRG payment is defined here as any per-case hospitalpayment method in which differences in case mix aretaken into account using DRGs to classify case types.

New Jersey is currently the only State in which allpatients and all hospitals are subject to DRG-specificrates per case. * Maryland currently uses a case-mixindex approach in about 30 of its 51 hospitals, and thecurrent Medicare hospital reimbursement system es-tablished by TEFRA sets maximum limits on per-casepayment using a DRG case-mix index. Georgia has ex-perimented with the use of DRGs to define hospitalgroups for per-case payment but is no longer using thesystem.

New Jersey began using the old 383 DRGs and isnow using the modified 467 DRGs in its payment sys-tem. The new Medicare system bases payment on thenew DRGs. A Medicare case-mix index was developedfor TEFRA using the 467 DRGs. Georgia’s experimentwith hospital groupings was based on the old 383DRGs. These programs are described in this section.

*The State of New York currently uses a DRG case-mix index in its rateset-ting program, but there the unit of payment is the inpatient day, not the case.Payment of per-diem rates creates incentives that are quite different from per-case payment. Consequently, the use of DRGs in New York is not discussedhere.

70

Several other States are using, and the AmericanHospital Association (AI-IA) once proposed as a modelsystem, per-case payment systems that do not explicit-ly adjust for case mix. Descriptions of selected per-casesystems are presented below as well.

DRG-Specific Rates Per Case:New Jersey*

In 1978, New Jersey passed a law mandating thegradual implementation of a per-case payment systemcovering all payers. A Hospital Rate-Setting Commis-sion was given the power to adopt an approach thatties payment rates directly to the patient’s DRG. (Muchof the developmental work for this ratesetting methodwas funded by a $3 million Federal Health Care Fi-nancing Administration (HCFA) grant to the NewJersey Department of Health.) In May 1980, 26hospitals began billing patients on a DRG-specific rateper case. By October 1982, all New Jersey hospitalshad been brought into the DRG system (22).

The DRG payment system works as follows: eachpatient is assigned to a specific DRG on discharge, andthe hospital is paid a previously specified rate for thatDRG. All classes of payers must pay the assigned rateto the hospital regardless of the actual amount of re-sources consumed in treating the patient, with the ex-ception of these “outlier” cases: patients for whom thelength of stay is unusually short or long relative to themean stay in the DRG; cases where the hospital stayends with death; when the DRG is a low-volume cat-egory; or when discharge is against medical advice.These outlier cases are paid according to the hospital’scharges, which are themselves controlled under a pre-existing ratesetting approach.**

The DRG rate assigned to a hospital is constructedfrom data on the hospital’s own costs as well as thoseof all other similar hospitals in the State.*** A hospital-specific preliminary cost base (PCB) is first establishedby taking the hospital’s actual expenditures in a baseyear (2 years before the rate year). This PCB includesdirect patient care costs, indirect (overhead) costs, al-lowances for the replacement of capital facilities, baddebt and charity care, and working capital. Only the

● A detailed description of the New Jersey DRG payment system is pro-vided in a separate working paper, ‘Using Diagnosis Related Groups (DRGs)in Hospital Payment: The New Jersey Experience, ” by Joanne E. Finley (22).

● *Even these cases would not be reimbursed on a cost-reimbursement basis.They would be paid the DRG rate plus a perdiem rate for each day beyondthe “trim point.”

● **Hospitals are classified in three categories: major teaching, minor teach-ing, and nonteaching.

71

direct care component of the PCB is assumed to varywith the DRG. The direct care costs are allocatedamong DRGs using formulas that presumably reflectactual resource use by patients in various DRGs. Forexample, nursing costs are allocated among the hos-pital’s DRGs according to the percentage of total pa-tient days in each DRG, while ancillary departmentcosts are allocated among DRGs on the basis of thepercentage of total department charges in each DRG.

Each hospital’s DRG-specific average direct patientcare cost computed as above is the basis for calcula-tion of a statewide average cost per DRG, which be-comes a standard DRG rate. The hospital’s rate be-comes a blend of the hospital’s own average direct carecost per DRG and the peer group average (or standard)DRG cost. The portion of each cost average that isused (i. e., the hospital’s own or the standard) variesacross DRGs depending on the amount of statewidevariance in the costs of treating patients within a givenDRG. If there is substantial variation in the costs oftreating a DRG, greater reliance is placed on the hos-pital’s own cost experience. The percentage of thestatewide standard cost used in setting rates rangesfrom a low of zero percent to a high of 100 percent,with most DRGs falling into the 40- to 75-percent range(62).

After DRG-specific direct care costs are estimated,hospital-based physician costs and overhead costs areadded, and the total is inflated to the rate year. Otherallowable costs (e.g., allowances for capital facilitiesand equipment and charity care) are calculated andallocated among DRGs on a percentage basis. Hos-pitals are then paid these final DRG-specific ratesthroughout the rate year.

Under this system, hospitals may keep any surplusachieved by reducing per-case costs, but beginning inthe 1982 rate year, a part of any surplus resulting fromincreasing admissions is taken back in the final recon-ciliation. Similarly, increases in costs per case must beabsorbed by the hospitals, but revenue losses due todecreases in admissions are moderated by a formulaat reconciliation.

In theory, this method of DRG price constructioncontains built-in annual adjustments to DRG ratesthrough changes in the base-year costs to reflect chang-ing levels of resource use. The hospital’s rate for a par-ticular DRG could change as a result of either changesin its own costs of providing services or statewide peergroup changes in the costs of treating the DRG. Therate facing a particular hospital can change even if itsown and the statewide peer-group average costs do notchange. For example, if the variance among patientsin the cost of treating cases in a particular DRG wereto decrease due to greater standardization of treatment

across the State, the rate in subsequent years wouldbe based more heavily on the statewide average costand less on the hospital’s own costs than in the previ-ous year. In practice, staff and budget limitations haveprecluded timely updating of the base year. The 1983DRG rates are still based on 1979 costs, with only in-flation factors changing in recent years (65). The Com-mission expects to update the base year to 1982 forthe 1984 rate year (96).

Changes in specific DRG rates are also possiblethrough an appeals process in which any interestedparty, be it a hospital, a payer, a patient, or the Com-mission itself, may request a review of a rate in oneor more DRG category if it believes it is offering serv-ices using new, more costly technology. As of Febru-ary 1983, however, only a few DRG appeals had beenfiled, and none had been completed (96).

Per-Case Payment With a DRGCase-Mix Index: TEFRA

In August 1982, Congress passed landmark legisla-tion, the Tax Equity and Fiscal Responsibility Act of1982 (TEFRA) (Public Law 97-248), which moved theentire Medicare system toward DRG payment begin-ning in October 1982. TEFRA made major revisionsin traditional cost-based reimbursement with the im-position of a hospital-specific maximum limit on theamount of reimbursable inpatient operating costs.

The new Medicare approach, which went into ef-fect in October 1982, has two key elements:

● For 3 years starting in October 1982, a hospital’sinpatient operating costs per case will have a “tar-get” rate of growth determined by the general rateof wage and price inflation in the hospital’s re-gion. If its operating costs per case are below thistarget rate it may keep 50 percent of the savings,up to 5 percent of the target rate. If the hospital’scosts exceed the target rate, it will receive only25 percent of its excess costs in the first 2 years,and none in the third.

● In no case can the hospital’s reimbursement ex-ceed a per-case limit on operating costs establishedby DHHS. The hospital’s new limit is 120 percent*of the mean cost per case for hospitals of the sametype. (Each hospital is categorized according toits bedsize and location. ) The limit is adjusted upor down by a DRG-based index of case mix foreach hospital.

Neither of these provisions puts any limit on capitalcosts (depreciation and interest), direct teaching ex-

● This limit will be reduced over the 3 years to 110 percent.

72

penses, or outpatient services. These remain “pass-through” items.

The DRG case-mix index for a particular hospitalhas been computed as the sum over all DRGs of thenumber of cases in the DRG times its national relativecost weight. The relative cost weights were constructedfrom a 1979 20-percent sample of Medicare inpatientclaims (the “MedPar” file) which contains data onhospitals’ charges and clinical information. The weightassigned to a particular DRG is the ratio of the averagecharge (adjusted and standardized) per case in thatDRG to the average charge per case across all DRGs(18). Although the amounts that hospitals charge pa-tients do not necessarily correspond to the cost oftreating patients (69), a study of the relationship be-tween hospitals’ overall 1979 DRG index value andtheir total 1979 inpatient operating costs revealed asimple correlation coefficient of 0.60 between the two(69). Further analysis has shown that a given percent-age difference in the case-mix index is met with roughlythe same percentage difference in operating costsamong hospitals (69).

TEFRA did not represent a wholesale abandonmentof cost-based reimbursement. For those hospitalswhose costs are below both the per-case limit and thetarget rate, reimbursement will be on the basis of cost,with a small incentive payment added. Over time,however, as the limit becomes tighter, a greater pro-portion of hospitals will find themselves with the per-case limits as real price constraints.

It was not clearly specified in the law how oftenHCFA was to update the hospital’s case-mix index.There appeared to be no plans to update the indexthroughout the life of TEFRA. Annual changes in thecase-mix index value to reflect changing case loads areconsidered unnecessary on the assumption that in theshort run, a hospital’s case mix is relatively stable andnot easily manipulated (see ref. 2). This decision un-derscores the temporary nature of TEFRA provisions,which will be phased out after 3 years as the new Med-icare law is implemented.

DRG-Specific Rates Per Case:The Medicare Law

In April 1983, the President signed into law a sweep-ing revision of Medicare’s inpatient hospital paymentsystem (Public Law 98-21). Begining in October 1983,the new payment method will evolve over a 3-yeartransition into a national set of DRG-specific pricesadjusted only for the hospital’s area wage rate and itsurban or rural location. DRG prices will apply to vir-tually all short-term acute-care general hospitals in theUnited States.

The new system will gradually supercede TEFRA,which moved the entire Medicare system from retro-spective cost-based reimbursement toward DRG pay-ment. The provisions of TEFRA (summarized above)did not represent complete abandonment of cost-basedreimbursement, but after the 3-year transition period,the new Medicare system will virtually replace retro-spective cost-based reimbursement with a prospectivepayment system based on DRG prices.

During the 3-year phase-in period, only part of thehospital’s payment will be on the basis of a DRG price;the remainder (a percentage decreasing each year) willbe made on the basis of its own reasonable costs (withmaximum limits as designated by TEFRA). Capitalcosts will continue to be paid for totally on a retrospec-tive cost basis* until the end of the transition period,at which time the law contemplates, but does not speci-fy the method for, the incorporation of payment forcapital into the DRG pricing system.

The pricing system will apply to all inpatient ad-missions except for a small number of cases (set as apercentage of the total by statute) with unusually longlengths of stay. The rate of payment for these caseswill be increased by the estimated incremental mar-ginal costs of care during the extended stay.

The initial national set of DRG prices will be basedon the 1980 average inpatient operating cost per casefor each DRG in a 20-percent sample of Medicare in-patient claims. The law requires that DRG prices beupdated regularly in two ways. First, an overall an-nual rate of increase is applied to every DRG to keeppace with the general level of inflation and rate of tech-nological change in the economy. Second, the relativeprices (i.e., the ratio of one price to another) must beassessed and recalibrated at least once every 4 years,with the first recalibration scheduled for October 1985.The recalibration process must reflect changes in treat-ment patterns, technology, and other factors that alterthe relative use of hospital resources among DRGs.The Prospective Payment Assessment Commission es-tablished by the law will be responsible both for mak-ing recommendations regarding recalibration and forevaluating any such adjustments made by the Secre-tary of DHHS.

Certain kinds of hospitals, such as long-term, psy-chiatric, and children’s hospitals, will be exemptedfrom DRG payment. Teaching hospitals are included,but for the present the direct costs of teaching (e.g.,residents’ and interns’ stipends) will be retrospective-ly reimbursed on the basis of cost, and a further ad-justment will be made for the indirect costs associatedwith teaching. In addition, the law requires the Med-

‘For-profit hospitals will be paid a return on equity as part of the capitalcost reimbursement.

73

icare program to participate in any State-legislatedalternative prospective payment program that coversat least 75 percent of the State’s population, makes pro-visions for competitive health plans, assures theFederal Government that access to hospital care forMedicare and Medicaid beneficiaries will not decline,and assures the Federal Government that hospital costsin the aggregate will be no higher under the State pro-gram. If the State system leads to hospital costs thatare higher than would be expected under DRG pay-ment, Medicare is empowered to recoup such over-payments from hospitals in subsequent years. Thus,States will probably move cautiously to adopt alter-native all-payer prospective payment systems.

The law also puts into place a mechanism for qualityassurance and utilization review by requiring hospitalsto contract with regional peer review organizations ata fixed price per review as a condition of payment.The payments for such reviews will come out of theMedicare Hospital Insurance Trust Fund and areguaranteed by statute.

DRG Case-Mix Adjustment: Maryland’sGuaranteed Inpatient Revenue System

The State of Maryland has been regulating hospitalrates since 1974, when hospitals’ charges were frozenpending the implementation of a new ratesetting ap-proach. From its inception until 1977, the Health Serv-ices Cost Review Commission was empowered to setrates for all payers except Medicare and Medicaid. In1977, a waiver from Medicare and Medicaid regula-tions was granted by HCFA; since then, the systemhas included all payers.

The ratesetting program has evolved over time withdifferent methods applied to different hospitals. TheGuaranteed Inpatient Revenue (GIR) method, whichuses DRGs, was first employed in 1976 in 14 hospitals(53). Today, approximately 30 of Maryland’s 51 hos-pitals are paid by this method, including all hospitalswith 400 beds or more.

Maryland’s GIR method is essentially a revenue con-trol system, where the allowed amount of revenue percase in each DRG is based on the hospital’s actualrevenue per case in the DRG in a selected base year.Hospitals do not bill the payer directly by the case;they bill on the basis of approved charges for eachservice provided. At the end of the year, the actualrevenue per case in each DRG is compared to the al-lowed revenue for that DRG. If actual revenue receivedper case exceeds the previously set allowed revenueper case in a DRG, implying that the intensity of serv-ices and/or length of stay have increased from the baseyear, the hospital’s revenue per case for the following

year is reduced by the amount of the difference. If ac-tual revenues per DRG case fall short of approved rev-enues for the DRG, then the hospital receives the dif-ference in an increase in the following year’s approvedrevenues (8). Hospitals may keep all savings from re-ductions in per-case costs, but a part of the revenueobtained from increases in admissions is recaptured bythe Commission. The hospital is also partly protectedfrom losses due to decreases in admissions.

The GIR system was modified in 1980 to bring in-terhospital comparisons into the computation of ap-proved revenues per case. Currently, all hospitals inthe State are classified by geographic area. For eachDRG for each payer category (i.e., Medicare, Med-icaid, Blue Cross, and others), an average charge forthe group is established. Each hospital’s actual chargeper DRG in each payer category is then compared tothis standard in the group. If, on the average, thehospital’s charge is higher than the group standard bymore than twice the allowable inflation rate, thehospital’s approved revenue is adjusted downward.For example, if the allowable rate of inflation is 6 per-cent, and the hospital’s DRG-specific charges are 15percent above the group standard, then the hospitalis allowed an inflation rate of only 3 percent (15 per-cent - [2 x 6 percent]) in its computation of allowedrevenue for that DRG.

A second modification is also under consideration.For all GIR hospitals, the Commission is movingtoward DRG-based reimbursement at the level of thegroup standard, as opposed to the hospitals’ own base-year revenue levels. If, for example, the hospital’s DRGrevenues are 10 percent higher than the standard acrossall DRGs, but are well above the standard, say by 40percent or so, for one or two DRGs, the Commissionwill move the hospital’s rate toward the 10 percentfigure on all DRGs. This may have the effect of pro-viding disincentives to the hospital to increase thevolume of cases in the more profitable DRGs.

Per-Case Payment With DRGs:Georgia’s Medicaid Program

Georgia used DRGs as part of its hospital groupingsystem in a 1981-82 Medicaid reimbursement experi-merit. ’ As in any grouping scheme, the underlying as-sumption was that similar hospitals with similar casemixes and service characteristics should consume sim-ilar amounts of resources per admission (80). Thegrouping was accomplished by using two data sets,one containing the number of patients in each of the

● Since Jan. 1, 1983, Georgia’s Medicaid program has been operating ona cost-based reimbursement system using 1980 costs plus an inflation factor.

74

original 383 DRGs and the other containing data on20 service characteristics (e.g., bed size, surgicalfacilities, diagnostic radiology, etc.). The case-mix andservice characteristics data sets yielded over 400 bitsof descriptive information which were then used togroup similar hospitals via a cluster analysis. In 1981,12 groups were formed ranging in size from 7 to 20hospitals. When the process was repeated for 1982,there were 11 groups, one with 5 hospitals and theothers with 10 to 20 (101). ’

After grouping the hospitals, the Medicaid programcompared the operating costs, excluding certain costssuch as malpractice premiums, depreciation on capital,and education. A group limit on costs for the next yearwas set based on 130 percent of the mean. It had beenestimated that 10 to 12 percent of the hospitals wouldbe outside their group limits.

Approximately 160 hospitals participated in thisproject. In the first year, 19 were outside the limit, andin the second year, 22 were higher than the limit. Thepenalty for being outside the group limit depended onhow the hospital’s base-year costs had compared tothe mean. Hospitals would lose the difference betweenthe allowable inflation rate and the percentage abovethe base-year mean, with the maximum penalty be-ing the allowable inflation rate. For example, if ahospital were 8 percent over the group limit and if theoverall allowable inflation rate were 10 percent, thathospital would have an allowable inflation rate of 2percent (101).

Per-=Case Payment Without DRGs:California Medicaid

From 1980 to the present, California’s Medicaid pro-gram (Medi-Cal) has operated under a per-case hos-pital payment system without an explicit adjustmentfor case mix. ** For each patient, hospitals have beenreimbursed the lowest of: 1) customary charges,2) Medicare reasonable costs, or 3) a maximum costper discharge (CPD) calculated in a fixed base year(generally fiscal year 1980). The CPD comprised thehospital’s own base period costs adjusted by an infla-tion index, growth in service intensity, and pass-through costs (including items such as depreciation,

● If the system had continued after Dec. 31, 1982, it was anticipated thatthe grouping would be an annual exercise until the groups stabilized. TheGeorgia Medicaid program staff had discussed changes in the procedure thatmight have led to stabilization (e.g., using the original 83 Major DiagnosticCategories with some partitioning based on whether or not surgery was per-formed) (101).

● *This method is gradually being phased out in favor of a new “SelectedProvider Contracting Program, ” under which hospitals will be awarded con-tracts on the basis of competitive bidding to provide inpatient care for Medi-Cal recipients. Except for emergency admissions, Medi-Cal will pay for caredelivered to its enrollees only at contract hospitals.

interest, utility costs, and malpractice premiums).Changes in the number of discharges from the baseyear were reflected in adjustments to the CPD limit.Beginning in October 1981, the program began to re-duce allowable fixed costs in hospitals with very lowoccupancy (below 55 percent), thus reducing the costper discharge limit in those hospitals as a penalty.

Hospitals have the right to appeal their CPD limitto the Department of Health. For example, if a hospitalwere to find that its Medicaid patient load shifted frommore routine cases to a high-cost load of, say, cardiacsurgery, the hospital would have recourse to the ap-peals process. Otherwise, such case-mix changes fromthe base year to any rate year would not be reflectedin the CPD limit.

Per-Case Payment Without DRGs:The AI-1A April 1982 Proposal

AHA proposed a Medicare payment system for in-patient care based on a prospective fixed rate perdischarge. Although beneficiary liabilities for de-ductibles and copayment would remain, hospitalswould be permitted to charge up to $1,000 per dis-charge in addition to the rate received from Medicareif they do not agree to accept the Medicare fixed priceas payment in full. Each hospital’s rate would be basedon its own allowable costs in a base year with adjust-ments for capital expenditures, compliance expendi-tures, return on equity, high Medicare and Medicaidvolume, and self-insurance against professional liabili-ty suits. These costs, with adjustments in a given year,would be divided by the number of Medicare dis-charges to obtain the initial rate. An inflation factorset by an independent panel of economists would bea forecast intended to reflect input price inflation,medical technology advances, and regional differences.The Medicare fixed rate would be computed by mul-tiplying the base rate by the inflation factor. Hospitalswould be paid this fixed rate per Medicare discharge.

In its proposal, AHA asserted that hospitals’ casemixes would not change in the short term, for whichthis program was intended, because of long-standingadmitting patterns, medical staff relationships, andhospital policies and procedures. The proposal calledfor an appeals process, however, which would haveallowed hospitals to appeal their rates because of in-creases in the complexity of their case mixes.

Per-Case Reimbursement Without DRGs:Idaho’s Medicaid Program

Since 1979, Idaho’s Medicaid program has had a per-case maximum limit on payment for Medicaid hospital

75

stays. The limit in any year is calculated on the basis rate. At present, one-third of Idaho’s 47 hospitals areof the hospital’s previous year’s audited costs per case subject to the per-case limit (81). However, since Med-with an adjustment for inflation. Hospitals are reim- icaid represents only about 5 percent of hospital ex-bursed the lower of billed charges, allowable costs, or penditures in Idaho, the program is not likely to havethe per-case limit (92). The per-case limit is implicitly had much impact on hospitals’ fiscal positions or re-adjusted for changes in case mix over time by the use havior.of the previous year’s costs in calculating each year’s

References

References

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12,

13,

14,

15.

16,

Abrams, H. L., “The Overutilization of X-Rays,”N. kkgl. J. Med. 300:1213,1979.American Hospital Association, “American Hos-pital Association Proposal: Medicare ProspectiveFixed Price Payment to Hospitals, ” April 1982.Averch, H., and Johnson, L. L., “Behavior of theFirm Under Regulatory Constraint,” Am. Ekon,Review 52 (December 1962):1052-1069.Bentley, J. D., and Butler, P. W., “Case-Mix Re-imbursement: Measures, Applications, Experi-ments,” Hospita/ FinanciaZ Managment, March1980.Berki, S. E., ‘The Design of Case-Based HospitalPayment Systems,” Medical Care 21(1):1, Jan-uary 1983.Butler, P. W., and Bentley, J. D., The DiseaseStaging Case Mix of a Sample of Teaching Hos-pitak: A Technical Report (Washington, D. C.:Association of American Medical Colleges, 1982).Caldwell, J. M., “Home Care: Utilizing ResourcesTo Develop Home Care,” Hospitak 56(21):68-72,Nov. 1, 1 9 8 2 .Cohen, H. A., Executive Director, Health Serv-ices Cost Review Commission, State ofMaryland, personal communication, Feb. 21,1983.Cohen, H. A., Executive Director, Health Serv-ices Cost Review Commission, State ofMaryland, statement to U.S. Congress, SenateCommittee on Finance, Subcommittee on Health,hearings on Prospective Payment Systems pro-posed by Secretary Schweiker, Feb. 2, 1983.Corn, R. F., “Quality Control of Hospital Dis-charge Data, ” Medical Care 18(4), April 1980.Cromwell, J., and Kanak, J., ‘The Effects of Pro-spective Reimbursement Programs on HospitalAdoption and Service Sharing,” Health Care Fi-nancing Review 4(2):67, December 1982.Demlo, L. K., Campbell, P. M., and Brown, S.S., “Reliability of Information Abstracted FromPatients’ Medical Records,” Medical Care 16(12),December 1978.Doremus, H. D., “DRGs May Be Raising FalseExpectations,” Hospitals Aug. 1, 1980.Duffy, D. J., Executive Vice President, SuburbanMedical Review Association, Inc., Kenilworth,N. J., personal communication, Apr. 28, 1983.Dunham, A. B., and Morone, J. A., DRG Evac-uation Volume III-A: Political Evaluation (Prince-ton, N. J.: Health Research and Educational Trustof New Jersey, January 1983).Eli Lilly & Co., Lilly HospitaZ Pharmacy Survey’82, 1982.

17<

18.19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

Evans, R., and Walker, H., “Information Theoryand the Analysis of Hospital Cost Structure, ”Canadian Journal of Ekonomics, No. 5, August1972.Fkderal Register, Sept. 30, 1982.Fetter, R. B., Shin, Y., Freeman, J. L., et al.,“Case-Mix Definition by Diagnosis-RelatedGroups,” supplement to MedicaZ Care 18(2),February 1980.Fineberg, H. V., “Clinical Chemistries: The HighCost of Low-Cost Diagnostic Tests,” in MedicalTechnology: The Culprit Behind Health CareCosts7 proceedings of the 1977 Sun Valley Forumon National Health, S. H. Altman and R. Blen-don (eds.), DHEW publication No. (PHS)79-3216(Rockville, Md.: Department of Health, Educa-tion, and Welfare, 1979).Fineberg, H. V., “Gastric Freezing—A Study ofDiffusion of a Medical Innovation,” in MedicalTechnology and the HeaZth Care System, reportby National Academy of Sciences, Institute ofMedicine, and National Research Council, As-sembly of Engineering, Committee on Technol-ogy and Health Care (Washington, D. C.: Na-tional Academy of Sciences, 1979).Finley, J. E., “Using Diagnosis Related Groups(DRGs) in Hospital Payment: The New Jersey Ex-perience, ” prepared for U.S. Congress, Office ofTechnology Assessment, unpublished draft,March 1983.Garg, M. L., Louis, D. Z., Gliebe, W. A., et al.,“Evaluating Inpatient Costs: The Staging Mech-anism,” MedicaZ C% 16(3):191-201, March 1978.Gonnella, J. S., and Goran, M. J., “Quality ofPatient Care—A Measurement of Change: TheStaging Concept,” Medical Care 13:467, 1975.Goran, M. J., “The Evolution of the PSRO Hos-pital Review System,” Medical Care 18(5) Sup-plement, 1979.Griner, P. F., and Glaser, R. J., “Misuse of Lab-oratory Tests and Diagnostic Procedures, ” N.I&l J. Med307(21):1336-1339, Nov. 18, 1982.Hadley, J., “Background Information on the Re-imbursement of Hospitals’ Teaching Expenses, ”working paper No. 1302-01, The Urban Institute,prepared for the U.S. Department of Health andHuman Services, Health Care Financing Admin-istration, contract No. 95-l? -97176/3-02,December 1979.I-Iadley, J., Juba, D., Swartz, K., et al., “Alter-native Methods of Developing a Relative ValueScale of Physicians’ Services: Year 1 Report,” re-port No. 307s-4, The Urban Institute, prepared

79

—

80

29.

30.

31.

32.

330

34.

35.

36.

37.

38.

39.

40.

for the U.S. Department of Health and HumanServices, Health Care Financing Administration,contract No. (HCFA) 500-81-0053, February1983.Hershey, N., “The Defensive Practice of Med-icine: Myth or Reality, ” Milhank Mere. Fund Q.50(1):69-98, January 1972.Horn, S. D., “Does Severity of Illness Make aDifference in Prospective Payment?” Health&eFtiancial Management, May 1983.Horn, S. D., “Measuring Severity of Illness:Comparisons Across Institutions,” American J.Public Health 73:25-31, January 1983.Horn, S. D., and Schumacher, D. N., “AcuteMyocardial Infarction: Measurement of Varia-tions in Charges, Length of Stay, and Mortali-ty,” mimeo, The Johns Hopkins University, Balti-more, Md., 1980.Horn, S. D., and Schumacher, D. N., “An Anal-ysis of Case-Mix Complexity Using InformationTheory and Diagnostic Related Grouping,” kfec$ical Care 17(4), April 1979.Horn, S. D., and Schumacher, D. N., “Compar-ing Classification Methods: Measurement of Vari-ations in Charges, Length of Stay, and Mortali-ty,” Medical Care 20(5), May 1982.Horn, S. D., and Sharkey, P. D., “MeasuringSeverity of Illness: Predicting Hospital TotalCharges, Length of Stay, and Ancillary Charges,”technical report, Center for Hospital Finance andManagement, The Johns Hopkins University,Baltimore, Md., 1983.Horn, S. D., Sharkey, P. D., and Bertram, D.A., “Measuring Severity of Illness: HomogeneousCase Mix Groups,” Medical Care 21(1):14-30,January 1983.Hombrook, M. C., “HCUP Research Note 3: Re-gional Differences in Length of Hospital Stay, ”prepared as part of the Hospital Cost and Utiliza-tion Project for the Department of Health andHuman Services, National Center for HealthServices Research, unpublished draft, Feb. 3,1983.Hombrook, M. C., “Hospital Case Mix: Its Defi-nition, Measurement and Use: Part II. Review ofAlternative Measures,” Medical Care Review39(2):73-123, summer 1982.Hughes, S. L., “Home Health Monitoring: Ensur-ing Quality in Home Care Services, ” Hospitak56(21):74-80, NO V. 1, 1982.Jonsson, E., and Mark4, L. A., “CAT Scanners:The Swedish Experience,” HeaZth Care Manage-ment Review 2:37, 1977.

41.

42.

43.

44,

454

46.

47.

48.

49.

50.

51.

52.

53.

54.

55.

Joskow, P. L., Controlling Hospital Costs: TheRole of Government Regulation (Cambridge,Mass.: MIT Press, 1981).Keane, J., Health Services Cost Review Com-mission, State of Maryland, personal communi-cation, May 5, 1983.Kind, A. C., Williams, D. N., Persons, G., et al.,“Intravenous Antibiotic Therapy at Home,”Arch. of W. Med. 139:413-15, April 1979.Klastorin, T. D., and Watts, C. A., “On theMeasurement of Hospital Case-Mix,” MedicalCare 18(6), June 1980.Lave, J., and Lave, L., ‘The Extent of Role Dif-ferentiation Among Hospitals,” Health %rvicesResearch No. 6, spring 1971.Lee, R. H., “Not-for-Profit Firm’s Demand for In-puts Under Alternative Systems of Revenue Regu-lation: Hospitals’ Response to Cost-Based andProspective Reimbursement” (app. E), in A Studyof the Impact of Reimbursement Strategies on theDiffusion of Medical T~hnologies, vol. II, pre-pared for The Urban Institute, Health CareFinancing Administration grant No. 18-P-97113/3411, June 1982.Levitan, S. J., and Kornfeld, D. S., “Clinical andCost Benefits of Liaison Psychiatry,” Am. J,Psychiat. 138(6):790, June 1981.Lichtig, L. K., “Data Systems for Case Mix,” Top-ics in Health Care Financing, summer 1982.Lichtig, L. K., “’Medical Record Data and Uni-form Billing,” J. of AMRA, October 1981.Link, C. R., Long, S. H., and Settle, R. F., “CostSharing, Supplementary Insurance, and HealthServices Utilization Among the Medicare Elder-ly,” Health Care Ffiancing Review 2:25-31, fall1980.Lundberg, C. J., “Home Care Legislation: NewTax Bill Mandates Changes for HHAs,” Hospitak56(21):81-84, Nov. 1, 1982.Maloney, T. W., and Rogers, D. E., “MedicalTechnology-A Different View of the Conten-tious Debate Over Costs, ” N. Ehgl. J. Med.301(26):1413-1445, Dec. 24/31, 1979.Maryland Hospital Education Institute, Guide toRate Review in Maryland Hospitals, Vol. II(Lutherville, Md.: Maryland Hospital EducationInstitute, 1979).McNamara, E., “Home Care: Hospitals Redis-cover Comprehensive Home Care, ” Hospitals56(21):60-66, Nov. 1, 1982.Mitchell, J. B., and Cromwell, J., “AlternativeMethods for Describing Physician Services Per-formed and Billed” (executive summary), pre-

81

56.

57.

58.

59.

60.

61.

62.

63.

64.

65.

66.

67.

68,

69,

pared for Health Care Financing Administration,contract No. 500-81-0054, Oct. 12, 1982.Myers, L. P., and Schroeder, S. A., “PhysicianUse of Services for the Hospitalized Patient: AReview, With Implications for Cost Contain-ment,” Milbank Mere. Fired Q. 59:481-507, 1981.National Academy of Sciences, Institute of Med-icine, Reliabih”ty of Hospital Dkcharge Abstracts,publication No. IOM-77-01, (Washington, D. C.:National Academy of Sciences, February 1977).National Academy of Sciences, Institute of Med-icine, Reliability of Medicare Hospital DischargeRecor&, publication No. IOM-77-05 (Washing-ton, D. C.: National Academy of Sciences,November 1977).National Academy of Sciences, Institute of Med-icine, Reliability of National Hospital DischargeSurvey Data, publication No. IOM-80-82 (Wash-ington, D. C.: National Academy of Sciences,1980).National Academy of Sciences, Institute of Med-icine, Beyond Malpractice: Compensation forMedical Injuries (Washington, D. C.: NationalAcademy of Sciences, March 1978).National Association of Insurance Commission-ers, IWpractice Claims Final Compilation (Med-ical Malpractice Closed CIaims, 1975-1978) 2(2),September 1980.New Jersey Department of Health, Health Eco-nomics Service, “1982 Schedule of Rates,”mimeograph, June 1, 1982.New Jersey Department of Health, Health Eco-nomics Services, “DRG Appeals Process,” unpub-lished draft, Mar. 24, 1983.New Jersey Department of Health and HSA, Re-gional Health Planning Council, Application Ab-stract and Record, State Application No. 821108-07-01, HSA Application No. 82-101, Dec. 15,1982.New Jersey Health Research and EducationalTrust staff, personal communication, Feb. 18,1983.O’Brien, J. P., “Emerging Malpractice Trends: Ex-pansion of Hospital Liability Continues,” Hos-pitals 57(8):59-63, Apr. 16, 1983.Paxton, H. T., Carlova, J., Eisenberg, H., et al.,‘l%king Your Practice More Malpractice Proof,”Medica] lkonomics, Sept. 30, 1974, pp. 69-130.Perspective, “PSRO: Is Peer Review Dead? OrJust Federal Funding?” 16(4):13-18, winter 1981.Pettengill, J., and Vertrees, J., “Reliability andValidity in Hospital Case-Mix Measurement, ”Health Care Financing Review 4(2):101-127,December 1982.

70,

71.

72.

73,

74.

75.

76.

77.

78.

79.

80.

81.

82.

83.

84.

85.

Phillip, P. J., and Iyer, R., “Classification ofCommunity Hospitals,” Health Services Re-search, winter 1975.Richards, G., “Dial RM for Investigation: RiskManagers Take on the Police Work When ClaimsArise,” Hospitals 57(8):64-68, Apr. 16, 1983.Romeo, A. J., Wagner, J. L., and Lee, R. H.,“Prospective Hospital Reimbursement and theDiffusion of New Hospital Technologies,” pre-pared for The Urban Institute, 1982.Rosenbloom, R. S., “Technological Innovationin Firms and Industries: An Assessment of theState of the Art,” Technological hmovation: AGiticaZReview of Current KnowZe&e, P. Kellyand M. Kranzberg (eds. ) (San Francisco: SanFrancisco Press, 1978).Russell, L. B., Technology in Hospitals: MedicalAdvances and Their Diffision (Washington,D. C.: The Brookings Institution, 1979).Ryan, H. R., Partner, Deloitte, Haskins & Sells,testimony presented before U.S. Congress, SenateCommittee on Finance, Subcommittee on Health,hearings on the Medicare Prospective PaymentProposal, Feb. 17, 1983.Simons, S., statement of The American MedicalRecord Association before U.S. Congress, SenateCommittee on Finance, Feb. 17, 1983.Simborg, D. W., “DRG Creep: A New Hospital-Acquired Disease,” N. E@. ~. Med. 304(26):1602-1604, June 25, 1981.Simborg, D. W., “The Changing Clinical Char-acteristics of DRG Creep, ” unpublished paper,1983.Smits, H. L., “The PSRO in Perspective, ” N.Ehgl. J Med. 305(5):253-259, July 30, 1981.State of Georgia, Department of Medical Assist-ance, “An Alternative Reimbursement System forGeorgia Hospitals, Annual Report, 1980.”State of Idaho, Idaho Medical Assistance Pro-gram staff, personal communication, Feb. 21,1983.Stiver, H. G., et al., “Intravenous AntibioticTherapy at Home,” Ann. Int. Med. 89 (Part1):690-93, 1978.Stolar, M. H., “National Survey of Selected Hos-pital Pharmacy Practices,” AJHP 33:225-30,March 1976.

/

Thomson, J. L. G., “Cost-Effectiveness of an EMIBrain Scanner: A Review of a Two-Year Experi-ence,” Health Trendb 9:16, 1977.Thompson, M. S., and King, C. P., “PhysicianPerceptions of Medical Malpractice Processes andDefensive Medicine,” unpublished paper, Sep-tember 1982.

82

86,

87,

88.

89.

90.

91.

92.

93.

94$

95,

U.S. Congress, Congressional Budget Office, TheEffkct of PSROs on Health Care Costs: CurrentFindings and Future Evaluations (Washington,D. C.: U.S. Government Printing Office, June1979).U.S. Congress, Congressional Budget Office, 71eImpact of PSROs on Health-Care Costs: Updateof CBO’S 1979 Evaluation (Washington, D. C.:U.S. Government Printing Office, January 1981).U.S. Congress, General Accounting Office, Problems With Evaluating the Cost Ef&tiveness ofProf&sional Standar& Review Organizations(Washington, D. C.: U.S. Government PrintingOffice, July 19, 1979).U.S. Congress, Office of Technology Assessment,“Variations in Hospital Length of Stay: Their Re-lationship to Health Outcomes,” unpublisheddraft prepared by M. R. Chassin, contract No.233-7210.0, 1983.U.S. Department of Health and Human Services,Health Care Financing Administration, “PSROEvaluation,” Health Care Financing: Researchand Demonstrations in Health Care Financing19801981, publication No. 03144 (Rockville,Md.: Department of Health and Human Services,1982.U.S. Department of Health and Human Services,Health Care Financing Administration, Report toCongress: Hospital Prospective Payment forMedicare (Rockville, Md.: Department of Healthand Human Services, December 1982).U.S. Department of Health and Human Services,Special Programs Branch, Division of AlternativeReimbursement Systems, “Inventory of the Alter-native State Hospital Reimbursement Plans (TitleXIX),” Sept. 30,1982.Vertrees, J., U.S. Department of Health and Hu-man Services, Office of Research, Health Care Fi-nancing Administration, personal communica-tion, Feb. 20, 1983.Wagner, J. L., “The Micro-Costing Method,” inAlternative Methods of LXveloping a RelativeValue Scale of Physicians’ Services: Year 1 Re-port, prepared for The Urban Institute, HealthCare Financing Administration contract No.500-81-0053, February 1983.Wagner, J. L., Krieger, M. J., Lee, R. H., et al.,“A Study of the Impact of Reimbursement Strat-egies on the Diffusion of Medical Technologies,Vol. I: Summary and Findings,” prepared for TheUrban Institute, 1982.

96.

97.

98.

99.

100.

101.

102.

103.

104.

105.

106.

107<

Warren, J., Executive Director, New Jersey Rate-Setting Commission, personal communication,Mar. 5, 1983.Warren, J., Executive Director, New Jersey Rate-Setting Commission, personal communication,May 1983.Wasserman, J., “Some Preliminary Results Fromthe Health Research and Educational Trust’s Eval-uation of New Jersey’s DRG System, ” mimeo-graph, May 1982.Williams, S. V., Levy, J. H., Kominski, G. F.,et al., “The Causes and Significance of Clinical-ly Inappropriate Patient Assignments in the Orig-inal DRG System, ” mimeograph, Oct. 23, 1981.Wood, W. R., Kobrinski, E. J., and Ament, R.P., “A Study To Develop and Test Measures ofCase Mix Complexity and Case Mix Severity forHospitals, ’’prepared for the Health Care Financ-ing Administration under contract No.500-78-0002, Sept. 19, 1979.Woodbury, D., Director of Hospital Reimburse-ment, State of Georgia, Department of MedicalAssistance, personal communication, Feb. 22,1983.Worthington, N., and Piro, P., “The Effects ofHospitals’ Rate Setting Programs on Volumes ofHospital Services,” Health CAe Fhancing Review4(2):47, December 1982.Yale University School of Organization and Man-agement, Health Systems Management Group,“The New ICD-9-CM Diagnosis Related Groups(DRGs) Classification Scheme User Manual,”prepared for the Health Care Financing Ad-ministration under grant Nos. 95 97499/1-01 and9597499/1-02, 1982.Young, W. W., “Hospital Case Mix MeasurementOverview,” in A Case Mti Approach to FimdingHealth Care, submitted to The Robert WoodJohnson Foundation, Nov. 7, 1982.Young, W. W., et al., “Patient Management Cat-egories Supplement, ” unpublished paper, Febru-ary 1983.Young, W. W., Swinkola, R. B., and Hutton, M.A., “Assessment of the AUTOGRP Patient Clas-sification System, ” Medikal Care 18(2), February1980.Young, W. W., Swinkola, R. B., and Zorn, D.M., “The Measurement of Hospital Case-Mix,”Medical Care 20(5), May 1982.

0