Technology and Health Care in an Era of Limits (1992)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Part TI Managing Care in the United States

~2 The Growth of Managed Care in the Private Sector Michael Soper and David Ferriss The private sector is moving rapidly into managed care.) Over the past two decades, profound changes have occurred in the way health care is organized and provided. The American health care delivery system has been transformed from one in which independent, fee-for-service medicine was dominant to one that is increasingly characterized by multiple health care delivery systems based on contractual relationships between managed care organizations and physician and institutional providers. This paper attempts to identify the factors responsible for this rapid movement toward managed care and to suggest consequences of this movement for the devel- opment and dissemination of medical technology. EMPLOYER-PROVIDED HEALTH CARE BENEFITS IN THE UNITED STATES Employee group health insurance has been an accepted employer-pro- vided benefit for some four decades in the United States (Feldman et al., Managed care: A term often used generically for all types of integrated delivery systems, such as HMOs and preferred provider organizations, implying that they "manage" the care received by consumers (in contrast to traditional fee-for-service care, which is "unmanaged"). The terms defined in footnotes are based in part on a working paper by Jonathan P. Weiner and Gregory de Lissovoy of the Johns Hopkins University School of Hygiene and Public Health, Baltimore, Maryland. 37

38 MICHAEL SOPER AND DAVID FERRISS 1989~. Since the end of World War II, an increasing number of employees and their dependents have been covered by group health insurance plans; in 1988, 153.3 million individuals were covered by some type of employer- related group health insurance (Health Insurance Association of America, 1990~. These benefits traditionally have been provided through an indemni- ty insurance program; however, alternative systems of prepaid health care delivery have existed for some time (Mayer and Mayer, 1985~. Although employer-provided health care benefits have been of signifi- cant value to employees, the cost of providing these benefits has become increasingly burdensome to employers, particularly in an environment char- acterized by global competition. In a global marketplace, U.S. employers find themselves at a distinct competitive disadvantage because the high cost of health care a cost of doing business in the United States-is not always a factor in the pricing schemes of competitors in other countries (Office of National Cost Estimates, 1990~. Even more alarming is the rapid rate at which health care costs have been increasing over the past two decades (Schieber, 19909. This rise is evident in the increasing proportion of the gross national product that is consumed by health care costs. A special concern of employers is that private-sector health care costs are inflating much more rapidly than those of the public sector, a fact that many employers and analysts attribute to cost shifting. Cost shifting occurs when a health care provider, for example, a hospi- tal, attempts to meet costs that are not fully covered by one purchaser of its services by "overcharging" other purchasers-in other words, shifting costs from one purchaser to another. To the extent that Medicare, as part of the Prospective Payment System, pays hospitals for services according to flat, diagnosis-related group (DRG) rates that do not cover a hospital's full costs for providing a service, the hospital must cover its unmet costs by increas- ing its charges to so-called passive payers those payers who are not in a position to negotiate specific reimbursement amounts. Many argue that the private sector is being "taxed" (or overcharged) to support Medicare through this type of cost shifting. In a similar fashion, managed care organizations negotiate reimbursement contracts with hospitals and physicians, a course of action that effectively removes these organizations from the ranks of the passive payers. This strategy protects those employers who provide em- ployee health care benefits through managed care organizations from fur- ther cost shifting, but it also results in a smaller number of passive payers onto whom providers can shift future costs. Traditionally, these passive payers have consisted primarily of indemnity health insurance companies. Corporate America is quite serious about getting control of medical costs in order to stay in business; it considers itself literally in a fight for

MANAGED CARE IN THE PRIVATE SECTOR 39 survival in an increasingly competitive global marketplace. This desire to contain medical costs is the major driving force behind the growth of man- aged care in the private sector. . . . . ~ the next section of this paper examines those developments in managed care that have made this rapid growth pos- sible. Prior to pursuing this theme, however, it is important to point out that managed care is not the only way to control health care benefit costs. Two other methods have been considered by employers: (1) relinquishing employer responsibility for the provision of employee health care benefits, and (2) sharing more of the costs of health care benefits with employees. Relinquishing responsibility for the provision of employee health care benefits by transferring current benefit costs to increased salary and wages is attractive to many employers, because the inflation of salaries and wages is much less than that of benefit costs. The overwhelming majority of employers, however, realize that this approach is not a realistic option. Employer-paid health care benefits are firmly established in American soci- ety and are viewed as more of a right than a benefit by most employees. Given the current size of the U.S. deficit, it is unlikely that Congress will solve the significant health care access problems that presently exist by relieving American business of this responsibility. It is more likely that Congress will mandate some form of employer-financed health care bene- fits for those who are employed but who are currently without health care coverage, and extend public financing only to those who are unemployed. The second method, sharing more of the costs of health care benefits with employees, has been extensively pursued by employers during the past decade. Although first-dollar, 100 percent coverage of health care costs by major employers was common in the past, it is rare today. This increased employee cost sharing has taken two forms: (1) employee contributions in the form of paycheck deductions and (2) the imposition of cost sharing at the time of receiving benefits. Employees now tend to shoulder, through paycheck deductions, a high- er proportion of the total premium cost of health insurance or its equivalent. This trend is particularly pronounced for the cost of dependent coverage. In addition, an increasing number of employers have introduced "cafeteria" benefit plans in which employees may choose what benefits they wish to purchase with a given number of employer-provided benefit dollars. Such plans are often an attempt by employers to help employees become more value-oriented purchasers of benefits, with hoped-for results that include increased awareness of the cost of employee benefits (in particular, health care benefits) and a decrease in the inflation rate of this cost. In addition to increasing employee payroll deductions for health care benefits, employers have also sought to impose additional cost sharing at the time health care benefits are received. This most commonly takes the -

40 MICHAEL SOPER AND DAVID FERRISS form of higher deductibles2 and higher levels of coinsurance.3 Deductibles of $200, $500, or more are now more the rule than the exception. To distribute the financial burden of higher deductibles more equitably, an increasing number of employers now set the deductible as a percentage of salary; thus, more highly compensated employees will pay a higher deduct- ible before health care benefits apply than will lower paid employees. A similar approach has been taken with coinsurance the percentage of actual health care costs that an employee pays after the deductible has been met. Current coinsurance levels are typically 20 percent and in some in- stances may be higher. In addition, employers have increased the maximum out-of-pocket expense to which an employee is subject. Increasingly, this amount is also a percentage of the employee's salary rather than a fixed dollar amount. The rationale for increasing employee out-of-pocket costs for health care services has been to give employees a significant personal incentive to consider carefully the costs as well as the benefits of these services. In increasing costs, employers would help their employees become more so- phisticated, value-driven consumers of health care services. Without such incentives, employers would continue to bear the brunt of higher costs that anse, at least in part, from unrestricted use of the "health care credit card" by providers; employees (and providers) would continue to be the sole beneficiaries of the employer's largesse. Although employers have pursued increased use of employee cost shar- ing for health care benefits, there are limits to the extent to which cost sharing can be imposed. High out-of-pocket costs may pose a significant barrier to access, thus inadvertently delaying needed treatment, which may result in poorer outcomes and higher health care costs. A second limitation comes from the intense dislike of employees for cost sharing and other "take-aways" from their health care benefits, a fact driven home by the willingness of organized labor to strike over this issue above all others. Overall, the use of employee cost sharing to contain health care costs has not been successful. The "smart shopper" theory has proved to be flawed with respect to health care; sick patients have neither the inclination nor the information to shop around for the least costly health care option. Consequently, the amount of cost sharing that employers can impose on 2Deductible: That amount of covered medical care expense that the insured individual must pay before insurance benefits become effective (e.g., $200 per individual, $500 per family per year a deductible may also be calculated on the basis of a percentage of an individual's annual income.) 3Coinsurance: That portion of the total covered medical expense paid by the insured indi- vidual after the deductible is met; coinsurance is usually expressed as a percentage of the total (e.g., 20 percent).

MANAGED CARE IN THE PRIVATE SECTOR 41 their employees has probably reached its limit, without appreciably slowing the health care inflation spiral. An approach to containing health care costs by means other than in- creased employee cost sharing is to provide health care services through an alternative health care delivery system that attempts to give more value for the dollar in the form of lower costs and improved quality. In addition, its proponents hope that such an alternative system can favorably affect the rate at which health care costs are inflating. This is where managed care organizations enter the picture. MANAGED CARE AND ITS GROWTH IN THE PRIVATE SECTOR Managed care organizations do three things: (1) they establish a health care delivery system composed of physicians, hospitals, and other health care providers; (2) they merge the role of third-party payer with the role of provider; and (3) they actively manage the health care delivery process. In this way, managed care organizations both provide and manage health care, rather than just pay for it. They are active, not passive, players. A major advantage of managed care organizations is that they signifi- cantly reduce the ability of providers to shift costs. Managed care organiza- tions have contracts in place with both physician and institutional providers that specify and establish reimbursement arrangements. These arrange- ments are subject to negotiation and marketplace forces, but they constitute a real constraint on provider charges. In the private sector, other con- straints on provider charges are few in number and weak in their effects. For example, when health care is covered by passive payers, providers can increase their revenues by increasing the intensity (volume) of the ser- vice provided. Extensive diagnostic testing, prolonged hospital stays, and frequent follow-up visits are all examples of increased intensity of services for which the demand is strongly influenced by provider recommendations. The marginal benefit of such increased intensity is uncertain at best; fortu- nately, any potential harm or risk, if costs are not a consideration, seems even more remote. Given the lack of harm of such practices, little has been done until recently to interfere with the way providers manage the delivery of health care services. Indeed, the utilization managements practices of managed care companies that limit the delivery of some of these services are the only significant constraints in the private sector. 4 Utilization management: The management of health resources so as to ensure that appro- priate care is provided. Utilization management activities may include precertification of specific diagnostic and therapeutic procedures, authorization and concurrent review of inpa- tient treatment, management of catastrophic illness and injury cases, and discharge planning.

42 MICHAEL SOPER AND DAVID FERRISS The more cost shifting that takes place, the greater the impetus for private-sector employers to move toward a managed care arrangement to avoid being the target of unrestricted increases in provider charges and intensity of services. This desire to avoid cost shifting is a significant impetus for the growth of managed care in the private sector; coupled with the desire of employers to eliminate inappropriate utilization of health care services, it has fueled an ever-increasing demand for managed care. Changes have occurred in the managed care industry over the past 20 years to meet the demand of employers who want to move their employees into managed care delivery models. Such changes have been necessary to spur the development of a sufficient number of managed care organizations (all major health insurers now sponsor one or more) and more flexible products that encourage employees and their dependents to give up their historical freedom to choose their physicians and hospitals from among all of those available. Although the concept of prepaid health care was well established prior to the 1970s, the Health Maintenance Organization Act of 1973 provided significant impetus to the health maintenance organization (HMo)5 move- ment. The act was important in three respects: (1) it provided federal grants and loans for the development of new HMOs; (2) it superseded exist- ing legislation in approximately 20 states, thus removing bamers to this type of health care delivery system; and (3) it provided an additional mea- sure of credibility for HMOs as well as mandating employers to offer HMOs to their employees when HMOs were available (Soper et al., 1990~. SHealth maintenance organization (HMO): A prepaid organized delivery system in which the organization and the primary care physicians assume some financial risk for the care provided to enrolled members. The term health maintenance organization was coined by Paul Ellwood for the Nixon administration in 1972. It constituted a renaming of two existing delivery models: prepaid group practices ("closed-panel" plans) and independent practice associations (IPAs, or "open-panel" plans). Currently there are four basic HMO models: Staff-model HMO: A type of HMO in which the majority of enrollees are cared for by physicians who are on the "staff" of the HMO. Although these physicians may be involved in risk-sharing arrangements, a majority of their income is usually derived from a fixed salary. Group-model HMO: A type of HMO in which a single large multispecialty group prac- tice is the sole (or major) source of care for HMO enrollees. Network-model HMO: A type of HMO in which a "network" of two or more existing group practices have contracted to care for the majority of patients enrolled in an HMO. A network-model HMO sometimes also contracts with individual providers in a fashion similar to the operation of an IPA-model HMO. Independent practice association or IPA-model HMO: An "open-panel" type of HMO in which individual physicians (or small group practices) contract to provide care to enrolled members. The primary care physicians may be paid by capitation or by fees for service, often with a "withhold" risk-sharing provision. Physicians who participate in IPA-model HMOs retain their right to treat non-HMO patients on a fee-for-service basis.

MANAGED CARE IN THE PRIVATE SECTOR 43 In 1973, when the HMO Act was passed by Congress, only about 2 percent of private-sector employees were receiving their health care through HMOs. It was hoped that the new legislation would impel rapid growth of the HMO concept of comprehensive benefits and limited cost sharing within a cost-effective system of health care delivery. Despite high hopes, howev- er, HMO growth during the 1970s was slow; other changes in the managed care model were necessary to bring about the eagerly anticipated growth of the concept. One of these changes was a greater opportunity for developing indepen- dent practice association (IPA) HMOs. At the beginning of the 1970s, virtually all HMOs were of the staff- or group-model type. As these HMOs became successful and attracted increasing numbers of patients who were seeking more comprehensive benefits, physicians in private practice became concerned. The IPA-model HMO began to be viewed as a way for physi- cians to compete with the staff- and group-model HMOs without abandon- ing the private practice of medicine. The development of IPA-model HMOs, which used the basic building blocks of mainstream medicine, allowed HMO growth to accelerate. This type of HMO avoided the problem of enormous cash outlays for building health care facilities and hiring physicians before sufficient members were enrolled. In the early 1980s, it became possible for for-profit companies to raise capital and invest in the development of IPA-model HMOs; this trend also facilitated the growth of the HMO movement. As a result of these developments, most of the growth of HMOs in the 1980s came from IPA-model health plans rather than from staff- and group- model plans. IPA-model HMO enrollment increased from 19 percent of total HMO enrollment to 41 percent during 1980-1990. In 1985-1986 in particular, IPA-model HMOs surged in popularity, increasing from 181 to 345 (InterStudy, 1991~. The second important factor contributing to the rapid growth of man- aged care in the private sector has been the development of point-of-service (poS)6 products, also referred to as open HMOs or open-ended HMOs. InterStudy, the Minneapolis-based HMO think tank founded by Paul Ell- wood, characterizes these health care benefit products as follows: (1) peo- ple enroll in an HMO; (2) the HMO permits them to receive services out- side of the HMO provider network without referral authorization but requires them to pay an additional deductible or a copayment, or both, to do so; (3) 6Point-of-service (POS) plan (also called an open-ended HMO or open HMO): A type of HMO in which the enrollees are not "locked in." Enrollees may leave the HMO and still have certain services covered. Such out-of-plan utilization is usually subject to a significant degree of cost sharing (e.g., deductibles and coinsurance) unlike those services delivered within the plan.

44 MICHAEL SOPER AND DA VID FERRISS coverage is offered under a financing mechanism similar to traditional in- demnity insurance and is available at any time service is desired; and (4) benefits for services received outside the HMO network are typically less comprehensive than the HMO benefit and usually include deductibles, co- payments, or coinsurance (InterStudy, 1991~. Behind the popularity of point-of-service products popularity that ex- tends both to employees and employers-is the fact that Americans like to have choices. Many Americans do not wish to give up entirely, in return for better benefits and lower out-of-pocket costs, the freedom to choose their own physicians and hospitals. The point-of-service product allows the employee to take advantage of the. ~.omnr~hf~nci~P Unfit ^~ ~= lava Or to ~ ~ ~<~ ~ DIVE ~ _ V~A_~4~ in ~1~ 1w W out-of-pocket costs associated with HMO membership while retaining the right to seek care outside the HMO system from any licensed provider in return for higher deductibles and coinsurance. Employees can try the HMO, but if they find it not to their liking, they are not locked in to using it. Point-of-service health care benefit products also have advantages for employers. Employers can offer their employees a single product that can operate either as a traditional HMO or as a traditional indemnity health insurance program. Employees who always use the HMO system receive all the benefits associated with membership in a traditional HMO; employ- ees who ignore the HMO and consistently seek care from other providers still have indemnity benefits; and employees who use the plan in both ways have increased flexibility in meeting their health care needs. Point-of-service products have also allowed employers with traditional indemnity programs to ease their employees into a more flexible managed care program without completely mandating that employees use only the HMO system of physician and hospital providers. The employer's hope, of course, is that employees will be attracted to HMO providers whose use ensures the most comprehensive benefits and least out-of-pocket cost. The benefits to the employer are better control of costs and an environment in which quality can be monitored and improved. As point-of-service prod- ucts become more acceptable to employees, employers are likely to increase significantly the cost to employees of the opt-out component of such plans. Doing so will preserve freedom of choice but attach such a high price to it that receiving care outside the system will not ha. ~ r~.nli~tir. tinting for Brat emolovees in most c. . rat aim ~- - - ~ ~ Thus, the widespread development of IPA-model health plans and the availability of point-of-service products are facilitating the growth of man- aged care in the private sector, the implications of which are significant for the U.S. health care delivery system. As managed care becomes the domi- nant model for providing health care benefits, there will be fewer and fewer people to whom costs can be shifted, and cost shifting will come to an end. When no one is buying "retail" anymore, the "wholesale" price will have to

MANAGED CARE IN THE PRIVATE SECTOR 45 be sufficient to cover the costs of providers of health care services. Less efficient providers that are unable to cover their costs will be forced to close. The most likely result will then be a far smaller number of more efficient, competing health care delivery systems rather than the thousands of independent physicians and dozens of redundant hospitals found today in most major urban areas. There will still be choices, but they will be fewer in number. IMPLICATIONS OF MANAGED CARE FOR TECHNOLOGY DEVELOPMENT AND DISSEMINATION If, indeed, managed care continues to grow in the private sector, what are the implications of this growth for the development and dissemination of medical technology? The remainder of this paper addresses this ques- tion, particularly in the light of Wennberg's argument (in this volume) in support of global limits on health care resources as opposed to managed care. In his paper, Wennberg argues for reform of the currently dominant delegated decision model in which the physician, without adequate knowl- edge of the shape of the benefit-utilization curve or the risk-benefit prefer- ences of the patient, makes decisions regarding what patients need and want with respect to medical care. Wennberg advocates a shared decision model based on outcomes research and a sensitivity to the desires of the patient. He contrasts this model with managed care, a model, he argues, that con- centrates on micromanaging physician decision making in the face of limit- ed knowledge and without considering the desires of the patient. Both models, in Wennberg's assessment, are unlikely to control the continuing escalation in health care costs. To achieve this objective, Wennberg advo- cates a health care system that imposes global limits on the availability of health care resources. Such limits, he claims, will eliminate excess capacity (and therefore reduce costs) without compromising access to care that is medically appropriate and desired by patients. Wennberg cites studies of geographic variation in hospital utilization as evidence that physician and hospital bed supplies are in excess of what is required to provide care that is both effective and desired by patients. Although it is true that managed care to a large extent has concentrated on reducing cost shifting, unnecessary hospital days, and inappropriate sur- gical procedures, it has also focused on improving the quality of medical care and the service that patients receive-that is, the outcomes of care, including patient satisfaction. This paper suggests that the managed care model has the greatest potential for bringing about both the transformation of the medical decision-making model and the appropriate allocation of health care resources that Wennberg advocates.

46 MICHAEL SOPER AND DAVID FERRISS Managed care organizations, because they are systems, have the means to incorporate into physician practice behavior the results of patient out- comes research and the new technologies that support patient involvement in the medical decision-making process. As Wagner points out (in this volume), managed care organizations, in particular, HMOs, have a unique opportunity, with their defined populations and provider panels and their management information systems, to serve as valuable laboratories for re- search in patient outcomes, cost-effectiveness, and technology assessment. Rather than interfering with the physician-patient relationship, managed care organizations have the potential to enhance the relationship in a way that brings value to both parties, in terms of higher quality and satisfaction as well as reduced costs. Wagner's example of how the Group Health Cooperative of Puget Sound, a staff-model HMO, made use of Wennberg's interactive computer video- disc technology to promote shared decision making between urologists and men with benign prostatic hyperplasia illustrates well the potential of man- aged care organizations to both improve the quality of medical care and contain costs. Although the incorporation of such technology in staff-mod- el HMOs is more easily accomplished than in other types of managed care systems (e.g., IPA-model HMOs), there appear to be no absolute barriers to the incorporation of this kind of technology in less tightly structured man- aged care organizations. With respect to excess physician and hospital bed capacity, it seems doubtful that the political will required to mandate resource ceilings will be found in the near future. In the past, certificate-of-need legislation has met with mixed success; the discipline of the marketplace may prove more ac- ceptable to providers and consumers of health care services alike. As the managed care sector continues to grow, an increasing volume of health care services will be provided through contracted provider systems- physicians, hospitals, ambulatory surgical centers, home health care agen- cies, laboratories, diagnostic imaging centers, and so on. Renewal of con- tracts by the managed care organization will increasingly depend on the ability of each provider to provide documented quality and service at a competitive price. In a competitive health care environment, providers who are unsuccessful at establishing themselves as high-quality, low-cost pro- viders will have excess capacity and consequently will not survive. A reduction in excess medical care capacity will thus occur as the result of competitive rather than political forces. The elimination of excess capacity in the managed care sector is further enhanced by the use of utilization management, procedures designed to assist the physician in determining which patients are the most appropriate candidates for a given clinical procedure and in identifying and coordinat- ing alternatives to hospitalization. To use Wennberg's illustration of the

MANAGED CARE IN THE PRIVATE SECTOR 47 black, white, and gray balls (see Chapter 2), managed care utilization man- agement procedures have the potential to help physicians distinguish those patients represented by the gray balls for whom there is strong reason to believe that a given surgical procedure or course of inpatient care will be beneficial, from other patients also represented by gray balls who do not require a given surgical procedure or inpatient care and for whom other modalities or sites of treatment are more appropriate. The role of managed care is to help physicians avoid using excess capacity just because it exists. By resisting irrational demands for such services, managed care organiza- tions alter the habits of physicians and hospitals that cause them to continue providing these services for no better reason than their availability. The effectiveness of utilization management is greatest for those cases (e.g., medical low back pain) that Wennberg identifies as showing wide, capacity dependent variation. A significant barrier to this approach, as Wennberg points out, has been inadequate knowledge of the benefit-utilization curve for specific clinical procedures; this is not a failure of managed care but simply inadequacy of collective clinical knowledge regarding the benefit of certain procedures and treatment modalities. Future advances in the study of patient outcomes, including functional health status and shared physician-patient decision making, will make it possible to develop utilization management procedures that are more scientifically based and that are oriented toward a physician education role rather than a policing mechanism. In the opinion of these authors, the impact of managed care on the development and dissemination of technology is significant. A primary fact is that technologies that provide both demonstrated benefit to the patient and reduction of costs will be readily adopted by managed care organiza- tions and readily disseminated through the health care system as increasing numbers of individuals are covered by managed health care arrangements. Technologies that are expensive and that have marginal or unproven benefit will be resisted by managed care organizations. The result will be in- creased pressure on the developers of a technology to demonstrate its bene- fit and its superiority over existing technology. Although the reluctance of managed care organizations to adopt a new technology before patient bene- fit has been clearly demonstrated may slow the process of new technology adoption and dissemination in the future, the longer-term benefit will be a more appropriate use of health care resources and additional assurance to the patient that a technology will be beneficial. The authors view this end result as a positive effect of managed care on technology development and dissemination. Griner, in a companion paper (in this volume), discusses the notion of "hidden technologies," those new applications and enhancements of a new technology that develop as a result of daily application in clinical settings

48 MICHAEL SOPER AND DAVID FERRISS following the technology's introduction. Such innovation is a natural, and sometimes desirable, consequence of adoption and dissemination; however, uncontrolled innovation is not always beneficial to patients and often serves only to increase costs without increasing benefits. In some cases, the inno- vation may even prove to be detrimental to the patient. Within traditional indemnity health insurance mechanisms, these hidden technologies are dif- ficult to recognize and control; with utilization management, managed care is able to resist their rapid adoption and to assess and introduce them in a more controlled manner, eventually adopting those that can be shown to provide increased patient value and rejecting those that do not. The manner in which managed care organizations currently evaluate new technology medical devices, pharmaceuticals and biologicals, and pro- cedures varies considerably from organization to organization. Small managed care organizations may have minimal or no internal capability for evaluat- ing new technology; consequently, they rely on external sources of technol- ogy assessment. Large managed care organizations, particularly the nation- al organizations, may have extensive internal resources for assessing new technologies and for making associated reimbursement decisions. Reim- bursement for a given technology may vary from organization to organiza- tion, but it appears that managed care organizations will attempt to base their decision of whether to adopt a given technology not on the basis of its potential to increase their market share but on a careful review of empirical data from methodologically sound evaluative studies. In essence, managed care organizations will attempt to base their technology adoption decisions (and thus reimbursement) on the true benefit (or lack of benefit) of a tech- nology. The scientific knowledge to support these decisions will not be sought or held for proprietary purposes Because of the complexity of the task of assessing new technologies, there is considerable interest in the United States in exploring ways in which managed care organizations, as well as traditional health insurers, might collectively support a level of technology assessment that is of higher quality than that currently possible for an individual organization to accom- plish. Representatives of four managed care and indemnity health insur- ance company trade associations are currently exploring a public-private solution to the problem of valid technology assessment. Achieving such a mechanism, within the limitations of antitrust legislation, has the potential to facilitate technology assessment that is in the best interest of the public and that will continue to encourage appropriate technological development and dissemination. Historically, managed care organizations and indemnity insurers have refused to pay for investigative or experimental technologies, arguing that funding such research is an appropriate role for government or private foun- dations and not for purchasers of health care benefits. This attitude, how

MANAGED CARE IN THE PRIVATE SECTOR 49 ever, has been increasingly challenged by both medical research institutions and by patients seeking coverage for investigative treatments in the absence of effective conventional therapies (Newcomer, 1990~. Largely as a result of increasing numbers of legal challenges to the policy of excluding investi- gative and experimental technologies from health insurance benefits, a number of managed care organizations and indemnity insurers have made excep- tions to their existing policies of denying coverage for unproven technolo- gies. In addition, U.S. Healthcare, a large regional HMO, and the Blue Cross/Blue Shield Association have announced their intent to support ran- domized clinical trials for evaluating the benefit of autologous bone marrow transplantation (ABMT) for women with breast cancer. Throb ~llthnrc Eli that m~nn~.d care organizations. i: 1111 ~ULll~l ~U~ll~ v ~L11~L 11101 ~_~ _ Ad, I,, , ~ _. n conjunction with traditional health care insurers, will be willing to support the evalua- tion of investigative and experimental technologies if uniform research pro- tocols are employed across multiple sites. Adherence to uniform protocols would allow collection of sufficient patient data to allow statistically signif- icant conclusions to be drawn. Currently, there is little uniformity, with the result that conclusions regarding the value of a given technology are de- layed and difficult. Such a practice serves only to confuse the evaluation of new technology and does a disservice to patients, for whose care a technol- ogy may be used that ultimately proves to be without benefit or even harm- ful. This paper concludes with several recommendations to help facilitate the development and dissemination of new technology within the dominant managed care environment of the future. 1. Managed care organizations should actively support outcomes re- search by making their data bases available to researchers in academic insti- tutions and by participating in well-designed studies. 2. Managed care organizations should support the evaluation of se- lected new technologies when studies are performed using uniform, meth odologically sound protocols. 3. Managed care organizations should refrain from competing for new technology on the basis of proprietary medical knowledge or exclusive cov- erage and instead should work cooperatively with the public sector to ad- vance the state of medical knowledge and valid technology assessment. 4. Managed care organizations, with their defined populations and provider networks, should take advantage of their potential for ongoing health services research in the areas of patient outcome measurement, shared physician-patient decision making, and evaluation of new therapies, and should freely disseminate the knowledge that results from such research.

so MICHAEL SOPER AND DAVID FERRISS REFERENCES Feldman, R., Dowd, B., Finch, M., and Cassou, S. 1989. Employer-Based Health Insurance. DHHS Pub. No. (PHS) 89-3434. Washington, D.C.: U.S. Department of Health and Human Services. Health Insurance Association of America. 1990. Source Book of Health Insurance Data: 1990. Washington, D.C.: The Association, p. 7. InterStudy. 1991. The InterStudy Edge: Managed Care-A Decade in Review, 1980-1990. Excelsior, Minn.: InterStudy. Mayer, T. R., and Mayer? G. G. 1985. HMOs: Origins and development. New England Journal of Medicine 312:590-594. Newcomer, L. N. 1990. Defining experimental therapy- a third-party payer's dilemma. New England Journal of Medicine 323: 1702-1704. Office of National Cost Estimates. 1990. National health expenditures, 1988. Health Care Finance Review 1 1(4): 1-41. Schieber, G. J. 1990. Health expenditures in major industrialized countries, 1960-87. Health Care Finance Review 1 1(4): 159-167. Soper, M. R., Stallmeyer, I. M., Bopp, K. D., and Wood, M. B. 1990. The HMO Act of 1973. In: E. J. Belzer, ed. Balancing the Triad: Cost Containment, Quality of Service, and Quality of Care in Managed Care Systems. Kansas City, Kan.: National Center for Managed Health Care Administration, pp. 12-18.

4 Managing Medical Practice: The Potential of HMOs Edward H. Wagner The traditional health maintenance organization (HMO), as defined by Donabedian (1983), remains an attractive paradigm for the provision of cost-effective health care. Prepayment imposes the discipline of a fixed budget, and responsibility for the full complement of services ensures a degree of balance between primary and specialty care. The presence of a defined population whose members have specific sources of primary care clarifies accountability, which permits epidemiologically based planning and management of primary and preventive care services. The successes of this model in reducing costs and hospitalizations without jeopardizing health status are reasonably well documented (Luft, 1981; Manning et al., 1984; Wagner and Bledsoe, 1990~. Interestingly, these successes were evident in traditional HMOs like the Group Health Cooperative of Puget Sound (GHC) long before the formal management of care (e.g., guidelines or preadmis- sion review) was in place. Strong, medically conservative cultures and the limitation of supply through controlled staff and bed expansion, analogous to the setting of global limits described by Wennberg (in this volume), would appear to be likely explanations. However, in the previous chapter, Soper and Ferriss point out that tradi- tional staff- and group-model HMOs have played only a small role in the recent rapid growth of capitated, managed health care systems. The new breed of HMOs are an array of acronymic insurance/delivery arrangements (Welch et al., 1990) characterized by capitation of some portion, often a 51

52 EDWARD H. WAGNER small one, of a physician's practice. Many modern HMOs are little more than financial arrangements without the culture and resources associated with truly shared responsibility by provider and insurer for the full spec- trum of services. In such organizations, managing care tends to rely on the "micromanagement" described by Wennberg. There is another approach to managing medical practice besides the imposition of rules or restraints on provider decision making (microman- agement) or the restricting of caDacitv hv settin~r ~rlohn1 limits {m~rnm~n_ agement). This approach to managing care refers to a collaborative process for planning and delivering health services to a population. Its objectives are to optimize health status and satisfaction with care and to do so at a reasonable cost and with acceptable provider satisfaction. Well-managed care includes active efforts to ensure that cost-effective services are re- ceived, rather than only intervening to reduce or eliminate ineffective ones. Care management, from this perspective, focuses on the needs, preferences, and outcomes of subpopulations of consumers categorized by age, condi- tion, or other relevant characteristics. Micromanaging care at the level of the individual patient through guidelines, reminders, or Recertification review is part of this perspective, but perhaps not the major part. A population perspec- tive forces attention to issues of supply, deployment of resources, and the organization of practice as well as "guidance" for individual patient decisions. It suggests an array of new intervention possibilities, such as the use of centralized resources for consumer and provider education, surveys of out- comes or preferences, and computer systems to create practice environ- ments more conducive to providing cc)st-effer.tiv~. ~.Z'r`~ {Wz'en~r earl Thompson, 1988). ~. _=, _ ~__ in ~45 ,~,~VV~1 111111 (lil~lulllall O ~' ' ~4-~^ BAAS Managing care to maximize outcomes requires collaboration among gen- eralists, specialists, consumers, administrators, and evaluation scientists. Collaboration is facilitated by a strong provider organization and culture that is willing to consider other issues besides professional autonomy and reimbursement. Organized consumer involvement provides a clearer role and voice for patients. A balance among these perspectives is usually healthy and nourishes the roles of science and data as adjudicators of differ- ences. Finally, managing care has often been a piecemeal operation a little quality improvement here, utilization review there, a guideline here, and so forth. Paul Ellwood (1988) has urged that health plans develop central nervous systems" to coordinate the various sensory and motor sub- systems that influence outcomes. Traditional HMOs have many of these ~ ~ . ~ elements in place to varying degrees, but uncoordinated and uneven use of them limits the ability of managed care to improve outcomes or reduce cost inflation.

HMOs AND MANAGING MEDICAL PRACTICE MANAGING CARE AT GHC 53 The way in which new technologies are handled at GHC, and by most group- or staff-model HMOs, illustrates both the problems and the potential of traditional HMOs (Figure 4-11. If a new idea surfaces, be it a preventive strategy, a drug, a device, or a significant change in protocol, it is referred to one of four standing committees: Prevention, Pharmacy and Therapeu- tics, Emerging Technologies, or Practice Efficacy (Guidelines). The most common sources of new ideas are GHC medical specialists who may be responding to a perceived community standard among their peers, to the clinical literature, or to the entreaties of a manufacturer. The committees, with varying degrees of clinical epidemiologic sophistication, review the literature and whatever local data might be available. An attractive, but expensive, new idea would require further review by management and gov- ernance structures responsible for coverage decisions. Ideally, implementa- tion would be followed by evaluation and reassessment, but all too often this has not been the case. The economic framework within which HMOs consider new technolo- gies is outlined in Figure 4-2. The rectangle represents the HMO's budget, which is largely fixed by its various capitation arrangements. Price compe New idea 1 Committee Review current practices and \ / outcomes \ / Make recommendation If costly, budget and benefits review 1 Implement l Review literature on innovation FIGURE 4-1 Clinical innovation management in HMOs. Evaluate . _ _

54 (B) (A) I(B) Fixed Budget (A) I (A) Arrows represent market forces: (A) = New Technologies/Quality of Care Concerns (B) = Price Competition/Lower Dues (B) FIGURE 4-2 Economics of innovation in an HMO. tition, a reality of the past decade, presses inward. New technologies, presumably to improve quality of care, push outward because most will be covered by the comprehensive benefit. EDWARD H. WAGNER (A) (B) BREAST CANCER SCREENING The development and implementation of an organized breast cancer screening program illustrate the process and logic of innovation manage- ment at GHC. In fairness, this example shows the HMO at its best. The process began in 1982, although the first organized services were not deliv-  ered until 1985, which suggests both the deliberateness of the process and the difficulties of decision making in large organizations. In 1982, GHC had no formal breast cancer screening program, but rising demand was overloading available mammography units. Fewer than 10 percent of the 110,000 adult women patients of the HMO had had a GHC mammogram.

HMOs AND MANAGING MEDICAL PRACTICE 55 Data from the Western Washington Cancer Surveillance System (CSS) showed that the incidence of breast cancer among GHC women and the distribution of their stages of disease at diagnosis had been stable for 20 years. These data confirmed that minimal effective screening was taking place. Growing provider and consumer pressure and an increasingly support- ive literature convinced GHC's prevention leaders that it was time for the Committee on Prevention (COP; Thompson et al., 1989) to review the is- sue. The COP reviews possible preventive interventions using explicit cri- teria based on the World Health Organization criteria for screening (Wilson and Junger, 19681: the disease or risk factor must be important, detectable, and responsive to treatment, and treatment must reduce morbidity and mor- tality. The COP criteria also explicitly include consideration of the feasi- bility of implementation at GHC and the costs relative to other competing budget demands. The COP appointed a breast cancer screening subcommittee of primary care physicians, epidemiologists, surgeons, radiologists, nurses, and health educators. In its addressing of the issue, the COP took a limited, budgetary perspective in assessing the economic impacts of various screening options on the HMO; in this instance, it attempted to estimate the costs of the additional screening associated with various screening program alternatives and the cost of evaluating those women with positive screening tests. Po- tential savings could occur if early-stage patients lived and paid dues long- er, and required less costly care. Carter and colleagues (1987) identified all GHC women diagnosed with breast cancer in 1972 and gathered cost data on all diagnostic and treatment services they received for breast cancer over the ensuing decade or until death. Care for women with earlier-stage dis- ease cost the system approximately $8,000 less than care for women with more advanced cancers. The prevailing screening guidelines in the community at the time were those of the American Cancer Society (ACS), which recommended baseline mammography and a clinical exam at age 35, with annual screening begin- ning at age 50. In addition to their earlier analyses, Carter and coworkers (1987) also estimated the future costs to GHC, over the next 5 years, of either doing nothing differently or of implementing the ACS recommenda- tions. According to their estimates, maintaining the status quo would not have generated sufficient screening activity to change the stage at diagnosis or the survival of many women, so it saved no money; on the cost side, it would require another mammography unit (at a cost of $231,000) simply to keep up with current demand. By contrast, implementation of the ACS guidelines would have substantially shifted the stage at diagnosis to earlier stages, added about 150 long-term survivors, and saved $2.6 million in breast cancer treatment costs and preserved dues revenues. Compliance with ACS guidelines, however, would have required new costs to the sys

56 TABLE 4-1 Risk Factor-Based Screening for Breast Cancer EDWARD H. WAGNER Risk Factors Risk Category Women 40-49 Women 50+ Screening Interval Women 40-49 Women 50+ Level 2 Level 1 Prior breast Same cancer, atypia, 2+ first-degree relatives One first degree relative Level 3 Level 4 Annually Annually One first- Every 2 Every 2 At least one . · . minor rlsK factora All other Not degree relative or 2+ minor risk factors a All other women women applicable years years Every 3 years On referral Every 3 years Not applicable aMinor risk factors include other family history of cancer, early menarche, late menopause, first birth after age 30 or nulliparity, or previous biopsy for benign disease. tem estimated at $6.8 million. million and the realm of possibility. The net cost to GHC exceeded both $4 In reviewing the breast cancer screening literature, the COP noted two points of interest: · Conventional risk factors for breast cancer could be used to stratify women into subgroups with very different levels of risks (see Table 4-1), level 1 having 4 to 14 times the risk of level 4 (Carter et al., 1987; Taplin et  al., 1990~. · The Swedish mammography trial (Tabar et al., 1985) screened wom- en every 2 to 3 years, rather than annually, and achieved reductions in mortality comparable to that achieved by the annual screening evaluated in the Health Insurance Plan randomized trial (Shapiro et al., 1985~. These observations suggested the possibility of a risk-based approach to screening. The COP considered a program in which a computer-scored risk factor questionnaire would establish a woman's level of risk, which would determine the interval between visits to a centralized screening cen- ter (Table 4-1~. At the center, women would receive a clinical breast exam- ination by a trained nurse-practitioner, a mammogram, and instruction in breast self-examination. Questionnaires and invitations to the screening center would be sent directly to women by the program rather than by physician referral. Primary care physicians would provide annual clinical breast exams and reinforce the centralized program. Cost estimates, using

HMOs AND MANAGING MEDICAL PRACTICE 57 the same methods as before, suggested shifts in the stage of disease at diagnosis and survival similar to the ACS approach, with the addition of women in their forties offsetting the losses associated with widened inter- vals between exams. Again, savings were estimated to be somewhat less than $3 million, but the estimated costs of screening were about the same (Carter et al., 1987~. This hypothetical zero net cost proved to be a crucial selling point despite a startup cost of more than $1 million. The COP sought approval but found no clear route. After a lengthy round of presentations to surgeons and radiologists, who feared patient overload; to primary care physicians, who feared loss of control; and to administra- tors, who were concerned about the budget, the Cooperative's Consumer Board of Trustees, led by its women members, finally approved the pro- gram and budget. The program, now in its sixth year, has steadily increased the propor- tion of women having at least one GHC mammogram (Figure 4-3~. Recent analyses of data from the Cancer Surveillance System suggest that the pro- gram has reduced the incidence of large and more advanced tumors among GHC women. Unlike other planned changes in medical practice, the Breast Cancer Screening Program is managed by a permanent steering committee and receives intense ongoing evaluation, supported in part by grant funds. The GHC breast cancer screening experience illustrates both the wis 50 40 30 cry a) 20 1982 1983 1984 1985 Year FIGURE 4-3 Percent of GHC women with GHC mammogram. 1986 1987 1988

58 EDWARD H. WAGNER dom of and difficulties with this approach to managing care. The intensive use of trained epidemiologists and explicit criteria helped build a defensible case for a creative program, which has been systematically reported in the peer-reviewed literature (Carter et al., 1987; Taplin et al., 1990~. Unfortu- nately, only the COP, of the four GHC committees, has that kind of evalua- tive science support. Formal, if limited, cost-outcome analysis was also essential to acceptance, and local data greatly strengthened the case. These analyses required laborious chart reviews, which could now be avoided by using computerized utilization and cost information. Consumer collabora- tion earlier in the process might have improved aspects of program design and increased program support as it went through budget deliberations. The major reason for the long development and decision process, however, was the absence of an organizational "central nervous system" with authority and accountability for changing medical practice. Instead, program devel- opers faced what became a trip through Wonderland to get action on their ideas. THE HMO AS EFFECTIVENESS LABORATORY In the case of breast cancer screening, published randomized trials sup- ported program development, but often this is not the case. In these latter situations, traditional HMOs have another opportunity- perhaps an obliga- tion to contribute more broadly to the effectiveness of medical practice by serving as laboratories for examinations of effectiveness and cost. All too often, careful review of the literature and available data do not support a clear recommendation about a new technology. The usual alternative is to leave it to clinical judgment, which, for new technologies, often means steady growth in use despite lack of formal review and approval. A second, better alternative is to conduct or participate in a formal study of a promis- ing new idea and to coordinate the study with the HMO's decision-making process for the issue in question. The randomized, controlled clinical trial (RCT) is the most persuasive basis for changing medical practice, a view now shared by practitioners as well as academicians. But the pace of progress by RCT is slow because of the constraints inherent in their use to answer urgent clinical questions. Given the manifold unanswered questions in clinical medicine and the gen- erally high cost of classic, academically based clinical trials, more expedi- tious, cost-efficient mechanisms are needed for mounting RCTs to evaluate the costs and benefits of new technologies. Health care systems, especially HMOs, offer important and underdeveloped research resources. Why should HMOs engage in the expensive, time-consuming, and, for most, very different business of conducting cost-effectiveness RCTs? Until recently, HMOs were far more interested in marketing surveys or evalua

HMOs AND MANAGING MEDICAL PRACTICE 59 lions of coverage changes and were reluctant to involve themselves in the complexities of studying care delivery. Now, however, health systems need to manage care delivery, and this requires the same kinds of evidence as are sought by the Health Care Financing Administration and the Agency for Health Care Policy and Research to meet national health policy objectives. Furthermore, managed care providers like HMOs now have information systems, patient populations, and research resources to gather the evidence for themselves. This creates an intriguing opportunity to accelerate the pace and relevance of outcomes or effectiveness research by encouraging research-oriented health systems to establish scientifically sound, locally relevant investigative activities to meet their own information needs as well as contribute to medical progress. How this kind of research scenario might work can be illustrated briefly with the interactive computer videodisc, a technology developed by John Wennberg and his colleagues to facilitate shared decision making. The Wennberg team sought GHC interest in serving as a pilot site for the benign prostatic hyperplasia (BPH) interactive program (Selikowitz et al., 1990~. GHC had no formal guidelines or review process for treatment of BPH; consequently, the HMO consulted its Committee on Practice Efficacy, GHC urologists, primary care leaders, and patient educators while deciding whether to participate in a pilot study. After further exposure to the idea and to a prototype, all agreed to the organization's participation in a pilot study, which was initiated in one of three GHC regions in November 1989. Initial physician and patient response have been very positive. Age- adjusted rates of prostatectomy for each of GHC's three regions before and after the initiation of the pilot were calculated using the HMO's computer- ized data systems. Rates dropped in the pilot region shortly after the initia- tion of the program and have remained stable in the other two regions. These differences persisted throughout 1990. The positive pilot experience provided the basis for a more formal test of the program. Under the leader- ship of Michael Barry at Massachusetts General Hospital, federal funding was obtained to conduct a formal RCT of the approach in GHC's other two regions. The objectives of this experiment are to compare the shared deci- sion-making program with a course of usual care, with respect to patient health and well-being, knowledge of their illness, willingness to participate in treatment decisions, and decisions that are finally made. CONCLUSION In the health care marketplace, the impact of the shift toward capitated and managed care on health care quality and costs remains uncertain. The heterogeneity of capitated arrangements and approaches to care manage- ment contributes to this uncertainty; it also militates against the opportunity

60 EDWARD H. WAGNER TABLE 4-2 System Characteristics of Health Maintenance Organizations Supportive of Managing Care 1. Defined population 2. Clear primary care accountability 3. Balance of power between primary and specialty care 4. Strong provider organization and culture 5. Organized consumer input 6. Access to scientific expertise and data 7. Structures and policies that limit capacity 8. Resources to support changes in practice style 9. Coordinated approach-"central nervous system" of capitated, managed care to improve care and reduce costs, particularly if attention is not directed toward identifying those aspects of managed health systems that are important to improving outcomes and reducing costs (Ell- wood, 19889. Table 4-2 suggests elements of HMOs that may be important for epidemiologically based management of care and for improving cost- effectiveness. Traditional staff- and group-model HMOs have many of these elements in place and remain important models for managing and researching medical practice, even if their potential in this area is still unrealized. REFERENCES Carter, A. P., Thompson, R. S., Bourdeau, R. V., et al. 1987. A clinically effective breast cancer screening program can be cost-effective, too. Preventive Medicine 16:19-34. Donabedian, A. 1983. The quality of care in a health maintenance organization: A personal view. Inquiry 20:218-222. Ellwood, P. M. 1988. Shattuck Lecture. Outcomes management: A technology of patient experience. New England Journal of Medicine 318: 1549- 1556. Luft, H. S. 1981. Health Maintenance Organizations: Dimensions of Performance. New York: John Wiley, pp. 185-207. Manning, W. G., Leibowitz, A., Goldberg, G. A., et al. 1984. A controlled trial of the effect of a prepaid group practice on use of services. New England Journal of Medicine 310:1505- 1510. Selikowitz, S. M., Albala, D. M., Barry, M. J., et al. 1990. Informed patient decision-making: Pilot project (abstract). Journal of Urology 4:413A. Shapiro, S., Venet, W., Strax, P., et al. 1985. Selection, follow-up and analysis in the Health Insurance Plan (HIP) Study. National Cancer Institute Monographs 67:65-74. Tabar, L., Fagerberg, C. J., Gad, A., et al. 1985. Reduction in mortality from breast cancer after mass screening with mammography. Lancet 8433:829-832. Taplin, S. H., Thompson, R. S., Schnitzer, F., et al. 1990. Revisions in the risk-based breast cancer screening program at Group Health Cooperative. Cancer 66:812-818. Thompson, R. S., Carter, A. P., and Taplin, S. H. 1989. Health promotion in an HMO: Ad astra per aspera. HMO Practice 3(3):82-88.

HMOs AND MANAGING MEDICAL PRACTICE 61 Wagner, E. H., and Bledsoe, T. 1990. The Rand health insurance experiment and HMOs. Medical Care 28(3):191-200. Wagner, E. H., and Thompson, R. S. 1988. Cancer prevention and HMOs. Cancer Investiga- tion 6:453-459. Welch, W. P., Hillman, A. L., and Pauly, M. V. 1990. Toward New Typologies for HMOs. Milbank Memorial Fund Quarterly 68(2):221-243. Wilson, J. M. G., and Junger, G. 1968. Principles and Practice of Screening for Disease. Public Health Paper No. 34. Geneva, Switzerland: World Health Organization.

5 Cost-Containment Efforts in the Public Sector: Oregon's Priority mist H. Gilbert Welch and Elliott S. Fisher This paper reviews the most publicized and most controversial portion of Oregon's plan for achieving universal health care coverage the priority list, released on February 20, 1991, ranking 714 condition-treatment pairs. It discusses the rationale and background for the effort to develop an ex- plicit and comprehensive prioritization of medical services, and it describes the methodology used to develop the list and the public involvement in the process. The paper also offers readers an opportunity to assess the face validity of the final rankings by presenting selected groups of condition- treatment pairs: the extreme high- and low-priority condition-treatment pairs, the priorities assigned to common conditions, the priorities assigned to the different indications for organ transplantation, and a random sam- pling of condition-treatment pairs. The final portion of the paper briefly discusses both the problems and advantages of public efforts to provide explicit limits on the provision of medical services. RATIONALE AND BACKGROUND To ensure that its entire population has health insurance, the Oregon legislature has proposed that all individuals with incomes above the federal poverty level have mandatory workplace coverage and that those with in- comes below the poverty level be covered through an expanded Medicaid program. The Medicaid plan uses both of the cost-containment strategies 63

64 H. GILBERT WELCH AND ELLIOTT S. FISHER outlined in Chapter 2: global limits and micromanagement. Global limits are imposed by heavy reliance on prepaid capitated care, a strategy that is being increasingly used by other state Medicaid programs (Welch, W. P., 1990~. The micromanagement scheme proposed in Oregon, however, is much more controversial and provides the focus for this discussion; it con- sists of a standard benefit package (for both Medicaid and the mandated workplace coverage) based on an explicit ranking of 714 condition-treat- ment pairs-better known as Oregon's health care priority list. The concept of a priority list grew out of a new paradigm of how costs should be contained in state Medicaid programs. The traditional method for dealing with budgetary constraints has been to rely on a combination of strategies: reduced fees to providers, simple limitations in the scope of optional services (such as eliminating dental services), and restrictions in eligibility. It has been assumed, however, that the federally mandated ben- efit package is fixed. A' '' ' ~11 _ _ _ 1 _ , . . ~ the result has been that the number of covered na~v~uua~s In most states fluctuates widely, with many individuals losing Medicaid coverage during economic downturns. Oregon proposes that ba- sic Medicaid coverage be guaranteed for the population with incomes below the federal poverty level; that is, the number of beneficiaries should be fixed and benefits should fluctuate instead. Thus, rather than eliminate individuals from the Medicaid rolls in times of budgetary shortfall, Oregon plans to eliminate services from the benefit package. To make such a paradigm operational requires a priority list of medical services. The process of change was initiated by the 1987 Oregon decision to curtail Medicaid coverage for organ transplantation (Welch, H. G., and Lar- son, 1988). Funding for bone marrow. heart liver anr1 nnn~r~.~ tr~n~niz~n_ ration was discontinued as part of an effort to provide revenue for an ex- panded program of basic health care. The trade-off was dramatically framed: the monies required to provide organ transplantation for a Droiected 34 , _ _, Ad-_ a-BAN ^-r^~^ patients In the biennium would purchase basic health care for 1,500 individ- uals who had not been covered previously. But the trade-off was also framed arbitrarily: the state targeted organ transplantation for elimination with little input from the public or from health professionals. Transplantation was chosen because it was an easily identified, new, and expensive therapy that had provided limited benefit (at that time, the 2-year survival rate for state-funded transplant recipients was less than 50 percent) to only a few people. There was little consideration of other health services as a potential source of savings. In 1989, the president of the Oregon state senate responded to these concerns by introducing new legislation. In 1989, Dr. John Kitzhaber authored Senate Bill 27 during the 65th Oregon Legislative Assembly to further expand coverage for basic services and to force a more open, rational, and comprehensive evaluation of the relative value of medical services. The bill established an 11-mem

PUBLIC-SECTOR COST CONTAINMENT: OREGON 65 her Health Services Commission (consisting of five physicians, represent- ing pediatrics, obstetrics, family practice, osteopathic medicine, and public health; four health care consumers; a public health nurse; and a social work- er) to oversee implementation of the plan (Welch, H. G., 19891. RANKING METHODOLOGY Although Senate Bill 27 specified that public input must be obtained, it left the methodology for establishing priorities to the commission. Public input was sought in three ways. First, public satisfaction with various health states was obtained from a random telephone survey of Oregonians. Half of the 2,000 individuals contacted were willing to complete the survey. Respondents were asked to rate 6 functional impairments and 23 symptoms. The average weights from the survey are shown in Table 5-1 and are adjusted to a scale of from 0 (death) to 1 (perfect health). The commission speculated that the most surprising result-that loss of consciousness (including fainting, seizures, and coma) was rated as a more satisfactory health state than the taking of prescription medication-was explained by the condition it is most often associated with in an unselected population a simple faint. Weights for the various health states were used in the net benefit formula described below. Second, the values ascribed by the public to various treatments (e.g., "treatment of conditions which are fatal and can't be cured the treatment will not extend the person's life for more than five years," "treatment for alcoholism or drug addiction") were determined during 47 public meetings held across the state. Oregon Health Decisions a community-based orga- nization that had surveyed the public in the past to identify the relative values placed on health care services (Crawshaw et al., 1990)-organized each meeting. More than a thousand Oregonians attended these meetings; two-thirds were health professionals, and fewer than 50 were Medicaid re- cipients. The participants were asked to classify each treatment into one of three groups: essential, very important, or important. The fundamental values that guided the foregoing classification were then identified and dis- cussed. The Health Services Commission later combined these values into three general attributes: value to society, value to an individual, and impor- tance to basic health care. The third approach for soliciting public input, which focused on specif- ic health services, was the sponsorship by the commission of seven public hearings. These hearings allowed special pleading by individuals and advo- cacy groups and eventually influenced the commission to raise the priority of a few selected services (e.g., preventive dental care). An initial ranking effort based on a strict, computerized cost-effective

66 H.GILBERT WELCH AND ELLIOTT S.FISHER TABLE 5-1 Weights Assigned to Various Health States Based on Responses from a Random Telephone Survey of 1,000 Oregonians Health State Functional Impairment Unable to drive In hospital/nursing home Limited in recreational activity Need help going to the bathroom or eating In wheelchair or walker under own control In bed or wheelchair controlled by someone else Symptom Wear glasses or contact lens Loss of consciousness (fainting, seizures, coma) Prescribed medication or diet Trouble talking Unable to stop worrying Overweight or facial acne Pain in ear/trouble hearing Trouble falling asleep or staying asleep Pain or discomfort in eyes Pain or weakness in back or joints Difficulty walking Severe fatigue/weakness Trouble with sexual performance Itchy rash over large area Pain with urination or bowel movement Headaches or dizziness Cough/wheezing/trouble breathing Drainage from sexual organs Often depressed or upset Trouble learning or remembering Stomach ache/vom~ting/diarrhea Severe burn over large area with pain Trouble with the use of alcohol or drugs - , ,- Weight .,, .954 .951 .938 .894 .627 .440 .945 .886 .877 .812 .785 .785 .783 .752 .752 .747 .747 .725 .724 .703 .701  .695 .682 .675 .674 .633 .630 .628 .545 Note: 0 = death; 1 = perfect health. The standard deviation of all weights is < .02. ness formula produced a preliminary priority list, which was released in May 1990. The list ranked 1,600 medical treatments, and it contained several serious flaws. Effectiveness and cost data were, of course, limited. But the major problem related more to the cost-effectiveness method itself, which seemed to favor minor treatments over life-saving ones. Widely criticized rankings included the priority of tooth capping over appendecto- my, reconstructive breast surgery over treatment for open fracture of the thigh, and treatment for crooked teeth over treatment for Hodgkin's disease.

PUBLIC-SECTOR COST CONTAINMENT: OREGON 67 Transplantation again was near the bottom of the list. Overall, there was widespread unhappiness with the ranking both inside and outside Oregon. In response, the commission adopted a new approach that modified the foregoing quantitative method by incorporating the consensus views of the commission. The final ranking was the result of three steps. First, to make the ranking more manageable, the commission chose to classify the list of condition-treatment pairs into 17 health service categories, which were then ranked relative to one another. Second, condition-treatment pairs were ranked within each health service category. Third, the 17 separate lists were combined, and those condition-treatment pairs that were judged by the commission to be out of order were reranked. In step 1, the 17 categories were ranked by summing the weighted category scores derived by each commission member. Before considering individual categories, each commissioner assigned a relative weight (out of a total of 100) to each of the three general attributes identified from the public meetings. The mean weights of the 11 commissioners were as fol- lows: value to society '10 (range: 20 to 60~; value to an individual-20 (range: 0 to 401; and importance to basic health care-40 (range: 20 to 50~. An individual commissioner's weight was held constant across categories. For each attribute, a commissioner assigned a value of from 1 to 10 to each of the 17 categories. A value of 1 and a value of 10 each had to be assigned to at least one category to ensure a distribution of scores. (A commissioner who had weighted "value to an individual" as only 5, for example, must still have assigned a value of 10 to at least one category for that attribute.) After the initial assignment, category values that were discovered to differ markedly from those of other commissioners could be modified. A summa- ry score for all commissioners was then calculated, which determined the ranking of the 17 categories (Table 5-2~. In step 2, the commissioners worked within categories. They first cal- culated the net benefit of treatment for each condition-treatment pair. The probability of different outcomes (e.g., death, trouble breathing, need for prescription medicine, return to former health state), both with and without treatment, were incorporated in the net benefit formula: Net Benefit = "outcomes with treatment x probabilities] toutcomes without treatment x probabilities] Probabilities for death, return to former health state, and intermediate out- comes were multiplied by a value for each health state. The value for death was 0; the value for return to a former health state was 1. Intermediate health states were assigned values obtained from the telephone survey (see Table 5-1~. The ranking of condition-treatment pairs within each category (step 2) was completed by acknowledging the cost of treatment and the number of

68 H. GILBERT WELCH AND ELLIOTT S. FISHER TABLE 5-2 Ranking by Members of the Oregon Health Services Commission of 17 Health Service Categories Rank Category Description (number of condition-treatment pairs in category) 3 4 s 6 7 8 9 10 11 12 13 14 15 16 17 1 Acute fatal treatment prevents death with full recovery (64) Maternity care (53) Acute fatal treatment prevents death without full recovery (61) Preventive care for children (4) Chronic fatal- treatment improves life span and quality of life (180) Reproductive/contraceptive services (4) Palliative care death is imminent (2) Preventive dental care (1) Proven effective preventive care for adults (3) Acute nonfatal treatment allows return to previous health state (60) Chronic nonfatal one-time treatment improves quality of life (107) Acute nonfatal treatment without return to previous health state (28) Chronic nonfatal repetitive treatment improves quality of life (80) Acute nonfatal treatment expedites recovery from self-lim~ted conditions (31) Infertility services (4) Less effective preventive care for adults (1) Fatal or nonfatal treatment confers minimal or no improvement in quality of life (31) individuals who would potentially benefit from the treatment. The commis- sion reordered condition-treatment pairs within a category to account for these factors. However, they also noted the difficulty of obtaining cost data in general, the limited ability to determine which costs could reasonably be attributed to treatment of a condition, and the virtual absence of data on the costs associated with no treatment. The minimal impact of cost data was demonstrated by a correlation analysis performed after the final ranking, which showed no correlation between cost and final rank. In step 3, the 17 separate priority lists (one for each category) were entered as blocks into a single document based on their category rank (i.e., all condition-treatment pairs in category 1 were placed on top, followed by those in category 2, etc.~. Selected condition-treatment pairs were then reor- dered "to reflect the clinical and public policy judgement of the Commis- sion" (Oregon Health Services Commission, 1991~. RESULTS The commission finalized the priority list on February 20, 1991. More than 60 pages long, it contains 714 condition-treatment pairs. Each entry includes diagnosis, treatment, the ICD-9 (the World Health Organization's International Classification of Disease, 9th ea.) coders) for the listed diag

PUBLIC-SECTOR COST CONTAINMENT: OREGON TABLE 5-3 Condition-Treatment Pairs with Extreme Rankings Rank Category Condition Treatment 69 1 1 Pneumococcal and other bacterial pneumonia Medical 2 5 Tuberculosis Medical 3 1 Pentonii~s Medical/surgical 4 1 Foreign body in airways or esophagus Removal 5 1 Appendicitis Appendectomy 710 17 Constitutional aplastic anemias Medical 711 11 Prolapsed urethral mucosa Surgical 712 17 Central retinal artery occlusions Paracentesis of aqueous humor 713 17 Extremely low birthweight (< 500 grams) Life support and < 23 weeks' gestation 714 17 Anencephaly Life support aThose cases for which treatment confers minimal or no improvement of quality of life. TABLE 5-4 Ranking for Common Condition-Treatment Pairs Rank Category Condition Treatment 16 48 144 176 292 305 317 318 337 393 408 11 478 502 586 685 9 s 5 11 11 11 700 13 10 13 17 Gallstones with cholecystitis Acute myocardial infarction Streptococcal sore throat Angina pectons Non-life-threatening arrhythmias Chronic obstructive pulmonary disease Hyperplasia of prostate End-stage renal disease Cataract (adult) Osteoarthritis  Transient cerebral ischem~a Osteoarthritis Transient cerebral ischeniia Intervertebral disc disorders (back) Gallstones without cholecystitis 14 Acute upper respiratory infections Cholecystectomy Medical Medical Medical/surgical Pacemaker Medical Transurethral resection Dialysis Removal Arthroplasty Endarterectomy Medical Medical Laniinectomy Medical/ cholecystectomy Medical nosis, the CPT- 4 (Current Procedural Terminology, 4th ea.) coders) for the treatment or procedure, the category number, and the rank. The calculated net benefit influenced the final ranking (Pearson correlation coefficient = .47, p < 0.0001~. The rank of a condition-treatment pair was also strongly

70 H.GILBERT WELCH AND ELLIOTT S.FISHER TABLE 5-5 Ranking of the Various Indications for Solid Organ Transplantation Rank Condition Organ Transplanted . ~.. 311 End-stage renal disease Kidney 363 Biliary atresia Liver 364 Cirrhosis of liver no mention of alcohol Liver 365 Myocarditis or pulmonary hypertension Heart 366 Acute hepatic necrosis Liver 507 Diabetes and end-stage renal disease Pancreas/kidney 609 Alphal-antitrypsin deficiency Lung 612 Hepatic cancer Liver 695 Alcoholic liver disease Liver Note: All condition-treatment pairs are in category 5, with the exception of acute hepatic necrosis, which is in category 3. TABLE 5-6 Ranking of the Various Indications for Bone Marrow Transplantation and Alternative Therapies Condition BMT Alternative Rank Alternative Therapy Rank Hodgkin's disease 208 Chemotherapy/radiation 188 Acquired aphasic anemia 213 Medical 259 ALL (child) 243 Chemotherapy/radiation 235 ALL (adult) 243 Chemotherapy/radiation 307 Agranulocytosis 248 None N. a. Constitutional 306 Medical 180 aplastic anemia ANLL 310 Chemotherapy 517 Chronic leukemias 518 Chemotherapy/radiation 278 Non-Hodgkin's lymphoma 696 Chemotherapy/radiation 238 Note: Alternative therapies listed in boldface are ranked higher than bone marrow transplanta- tion for the listed indication. All condition-treatment pairs are in category 5. Abbreviations: BMT, bone marrow transplantation; ALL, acute lymphocytic leukemia; N.a., not applicable; ANLL, acute nonlymphocytic leukemia. influenced by the rank of its health service category (Pearson correlation coefficient = .83, p < 0.0001~. Tables 5-3 through 5-7 give the reader a feeling for the final ranking as well as an opportunity to consider its face validity. Table 5-3 shows the condition-treatment pairs ranked at both extremes (the five highest and the five lowest ranks). Antibiotic therapy for bacterial pneumonia tops the list;

PUBLIC-SECTOR COST CONTAINMENT: OREGON TABLE 5-7 Ten Randomly Selected Condition-Treatment Pairs 71 Rank Category Condition Treatment 86 3 Subarachnoid and intercerebral BUIT holes/ hemorrhage craniotomy 246 5 Pernicious anemia Medical 254 5 Opportunistic infections in Medical immunocomprom~sed hosts 273 5 Regional enteritis Medical/surgical 383 13 Cerebral palsy Medical 490 11 Ganglion of tendon or joint Excision 494 5 Histiocytosis Medical 520 5 Anomalies of the gallbladder, bile Medical/surgical ducts, and liver 674 14 Acute pharyngitis and laryngitis Medical 706 17 Cyst of the kidney (acquired) Medical/surgical life support for a newborn with anencephaly is at the bottom. Table 5-4 shows the rankings for common condition-treatment pairs. For the two conditions for which medical and surgical treatment are listed separately, surgical therapy has the higher rank. Arthroplasty (e.g., hip replacement) has higher priority than medical therapy for osteoarthritis; endarterectomy has higher priority than medical therapy for cerebral ischemia. The motivation for the priority list was, in part, the recognition that the decision to curtail funding for organ transplantation had been arbitrary. Tables 5-5 and 5-6 demonstrate that organ transplantation is no longer at the bottom of the list. Table 5-5 shows the rankings of various indications for solid organ transplantation. These ranks were, in general, lower than those for bone marrow transplantation. Table 5-6 demonstrates the level of detail used by the Health Services Commission for this single procedure. Nine indications for marrow transplantation are shown, as are eight alterna- tive treatment strategies. Five of the alternative treatments (boldface) have a higher rank than marrow transplantation. Finally, to offer the reader an unselected sample, Table 5-7 includes 10 condition-treatment pairs drawn at random. DISCUSSION The level of funding provided by the Oregon state legislature will ulti mately determine which condition-treatment pairs are covered. The result ing benefit package, once a federal waiver is approved, will apply to the Medicaid program and will also serve as the minimum benefit package to be offered under the mandated employer-based insurance. Despite the fact that

72 H. GILBERT WELCH ANI)ELLIOTTS.FISHER the results of the priority list may only apply to a segment of the popula- tion, the Oregon experience deserves scrutiny for its potential relevance to a more universal health insurance plan. The mandate of the Oregon Legislative Assembly to develop a "list of health services ranked from the most important to the. 1~.~t imn~rt~nt" is particularly daunting task. Because no prototype was available for guid- ance, the Health Services Commission was forced to break new ground. Thousands of volunteer hours were devoted to the process by both profes- sionals and ordinary citizens. The community meetings and telephone sur- veys represent a level of public input rarely solicited in health policy plan- ning. Any health professional who reviews this process will be impressed by the effort committed to the task. But they will be equally impressed by the problems facing this kind of micromanagement. Three are particularly noteworthy. The first is accura Adapting either cost-effectiveness analysis or a consensus approach to such a comprehensive task was bound to be fraught with problems ~_ :1 _ ^~ cy. _ c, . . ~ ~ ~ ~ ~ _,~, ~ . . For By c;ona~on-~rea~ment pairs there are few data on effectiveness, a prob- lem compounded by the difficulty of comparing dissimilar benefits. Where- as the extreme high and low ranks shown in Table 5-3 have legitimate face validity, the relative priorities of more closely ranked services are, with good reason, open to question. For example, the higher priority (rank = ;~) assigned to treatment for end-stage renal disease in comparison with treatment for cataracts (rank = 337; see Table 5-4) may engender vigorous debate. Furthermore, because a line must ultimately be drawn, the accuracy of adjacent ranks is relative. For example. in T~hl~. ~ ~ ic it ~r~r~rm~riatm try ~· , . . . .. ~ - ~~ ~-~~~~~c-~~~ ~ ~~~~ ~ - ~ -~ ^~ ~$-~w V Iuna nears transplantation tor myocarditis (rank = 365) and not fund liver transplantation for acute hepatic necrosis (rank = 366~? Because the avail- able data were limited and the methodologies not sufficiently mature, the commission was forced to make decisions that in retrospect appear arbitrary and may not be reproducible. More important is the problem of heterogeneity. Even if perfect data were available and a consensus about the rankings could be obtained, ex- ceptions would exist. Priorities that make sense for the population as a whole may fail miserably for the specific case. Although internists may be unhappy to learn that arthroplasty (rank = 393) is more valued than ibupro- fen (rank = 478) in the treatment of osteoarthritis, it may be defensible to rank definitive treatment over palliation. For a patient with mild pain or one with extreme surgical risk, however, it is farcical. In addition, the heterogeneity of patient preferences is relevant (Barry et al., 19889. The ranking of surgical procedures above their medical alternatives ignores those patients who may prefer less invasive therapy. Although it is easy for clinicians to imagine such exceptions, it is difficult for policymakers to incorporate them into a set of priorities.

PUBLIC-SECTOR COST CONTAINMENT: OREGON ~3 Finally, there is the problem of administration. Since the priorities are catalogued by diagnosis, physician discretion in the choice of diagnosis could have a dramatic impact on the rank assigned their therapy. Some consideration must be given to the problem of "gaming" by physicians, particularly if financial incentives to do so persist. Will cultures be re- quired to ensure that physicians are treating strep pharyngitis (rank = 144) and not the common cold (rank = 7009? How can cholecystectomies for symptomatic cholelithiasis (rank = 16) be distinguished from those for asymptomatic stones (rank = 6859? The incentives for providers to-inflate diagnoses will be powerful when rank determines whether reimbursement occurs. Furthermore, it is unclear how a priority list can deal with patients whose symptoms have yet to be diagnosed or whose diagnosis does not appear on the list. Recognition of these problems has led Oregon's legislative leaders to emphasize a strategy that combines the priority list with capitation. In essence, however, it is a strategy that combines micromanagement with global limits. The priority list will define the benefit package on which the capitated rate will be based. Physicians will be free to prescribe treatments outside the benefit package but with no increase in total reimbursement. This approach will go a long way toward solving the problem of heteroge- neity by allowing for a more flexible allocation process tailored to the individual patient. By addressing incentives directly, capitation should lessen the problem of gaming. How capitated health systems will actually imple- ment the priority list, however, remains unknown. Where capitation is im- practical (e.g., rural Oregon), the problems of heterogeneity and administra- tion are likely to be more severe. Clearly, regardless of reimbursement strategy, implementation of a priority list in clinical practice will be difficult. On the other hand, there are certain aspects of a priority list that would offer benefits to physicians. The current pressures for cost containment force the health care system and the nation to face the problem of which services to deliver and to whom. Societal guidance about these choices would be appreciated, for two major reasons. The first is to mitigate the problem of malpractice by establishing agreed-upon standards of care that recognize the current environment of limited resources. Second, and more importantly, a system of public priorities will provide the ethical foundation needed to curtail services with token or undetermined benefit. A public policy not to support extremely low birthweight infants (rank = 713) could serve as an important example. In addition, a priority list might benefit society as well. Such a list forces explicit decisions about how to use limited resources and encourages public scrutiny of the process employed to make those decisions. Current- ly, much of this decision making is underground, and resource allocation decisions vary considerably. Because decisions are neither standardized _

74 H. GILBERT WELCH AND ELLIOTT S. FISHER nor publicized, there is little opportunity for public input. Application of a priority list within a universal health insurance plan might mitigate such problems. The need to set limits in health care coverage is sufficiently acute to warrant a consideration of priority setting, despite its problems. Admitted- ly, many refinements are needed. A more comprehensive approach will require evaluation of more condition-treatment pairs. The RAND Corpora- tion, for example, defined 864 distinct indications for carotid endarterecto- my alone (Chassin et al., 1987~. The problem of whose set of utilities are used to define health states also needs attention. The tension between using an unselected population (which may have minimal symptom experience) and diseased populations (which may have a tendency to inflate those symptoms they would experience without treatment) will have to be addressed (Mul- ley, 1989~. More consideration should also be given to the measurement of cost. Eventually, incremental cost must be calculated with careful consider- ation given to what is chosen as alternative therapy (Detsky and Naglie, 1990~. And, of course, more and better data are needed. Finally, given the objectives of this volume, the relationship between priority setting and technological innovation requires exploration. It is possible that the process of establishing priorities may change the rate of technological advance; it is also possible that technological advances will drive new priorities. The latter is certainly the case for a single state, whose health policy can have only minimal impact on innovation in medi- cine. On the one hand, Oregon has acknowledged the potential impact of new technologies on the priorities it has already established and intends periodically to reconsider its rankings. On the other hand, a widely applied priority list could influence the rate of technological advance with re- search on low-priority conditions and treatments Reemphasized. Although the rate of technological innovation is largely determined by the resources that are available, explicit priorities may govern where innovation occurs. Explicit priority setting has an important contribution to make in deter- mining what health care services to purchase in an era of limited budgets. It is especially relevant to health systems working within global limits, such as Britain's National Health Service (see Chapter 6~. The efforts being made in Oregon ought to advance this process. The state's approach has and will continue to foster vigorous debate (Duggan, 1989; Gore, 1990; Relman, 1990; Schwartz and Aaron, 1990; Hollingsbaum, 1991~. The meth- odology used to create the priority list is likely to undergo extensive peer review. The ranks of individual condition-treatment pairs undoubtedly will be a source of considerable dispute, encouraging further efficacy and cost- effectiveness research. Finally, Oregon's efforts should encourage those who say such a method is unworkable to develop some alternative.

PUBLIC-SECTOR COST CONTAINMENT: OREGON ACKNOWLEDGMENTS 75 The authors appreciate the efforts of W. Pete Welch, Ph.D., and Jenni- fer Dixon, M.D., for their assistance in this work. In addition, they are particularly indebted to the staff of the Oregon Health Services Commission and the Oregon Senate President's Office who were exceptionally respon- sive to requests for information. REFERENCES Barry, M. J., Mulley, A. G. Jr., Fowler, F. J., and Wenuberg, J. E. 1988. Watchful waiting vs. immediate transurethral resection for symptomatic prostatism: The importance of pa- tients' preferences. Journal of the American Medical Association 259:3010-3017. Chassin, M. R., Kosecoff, J., Park, R. E., et al. 1987. Does inappropriate use explain geo- graphic variations in the use of health care services? A study of three procedures. Journal of the American Medical Association 258:2533-2537. Crawshaw, R., Garland, M., Hines, B., and Anderson, D. 1990. Developing principles for prudent health care allocation: The continuing Oregon experiment. Western Journal of Medicine 321:1261-1264. Detsky, A. S., and Naglie, I. G. 1990. A clinician's guide to cost-effectiveness analysis. Annals of Internal Medicine 113:147-154. Duggan, J. M. 1989. Resource allocation and bioethics. Lancet 1:772-773. Gore, A. 1990. Oregon's bold mistake. Academic Medicine 65:634-635. Hollingsbaum, F. 1991. Rationing health care. British Medical Journal 302:288-289. Mulley, A. G. Jr. 1989. Assessing patients' utilities: Can the ends justify the means? Medical Care 27: S269-S281. Oregon Health Services Commission. 1991. Prioritized Health Services List of February 20, 1991. Salem, Ore.: The Commission. Relman,A. 1990. The trouble withrationing. New England Journalof Medicine 323:911- 912. Schwartz, W., and Aaron, H. 1990. The Achilles heel of health care rationing. New York Times, July 9. Welch, H. G. 1989. Health care tickets for the uninsured first class, coach, or standby? New England Journal of Medicine 321:1261-1264. Welch, H. G., and Larson, E. B. 1988. Dealing withlimited resources: The Oregon decision to curtail funding for organ transplantation. New England Journal of Medicine 319:171- 173. Welch, W. P. 1990. Giving physicians incentives to contain costs under Medicaid. Health Care Financing Review 12(Winter):103-112.