~ Pergamon PII: S0277-9536(96)00395-4

Soc. Sci. Med. Vol. 45, No. 4, pp. 545 554, 1997 ~E) 1997 Elsevier Science Ltd

All rights reserved. Printed in Great Britain 0277-9536/97 $17.00 + 0.00

THE OREGON EXPERIMENT: THE ROLE OF COST- BENEFIT ANALYSIS IN THE ALLOCATION OF MEDICAID

FUNDS

J A M E S F. B L U M S T E I N *

Vanderbilt Law School, Health Policy Center, Vanderbilt Institute for Public Policy Studies, 21st Avenue S. (a, Grand, Nashville, TN 37240, U.S.A.

Abstract--The state of Oregon decided to cover all potentially eligible Medicaid citizens to 100% of poverty. Previously, Oregon had covered persons up to 67% of poverty. In order to keep overall pro- gram costs in check, Oregon decided to limit the number of services that its Medicaid program would cover. Oregon's normative choice was to contain program costs by covering all eligible persons up to 100% of poverty, while at the same time uniformly limiting access to certain services for everyone in the overall group of eligible persons. The state developed a prioritization list of medical services and priced the components on the list. The amount of money ultimately available for the Medicaid program was a political decision informed by data about the cost of different services and influenced by the pri- orities set through an independent process of priority-setting. Physicians were asked to determine what works medically, how well it works, and what benefits accrue to patients. Recognizing that physician perspectives on efficacy might vary from patients' perspectives on valuation of benefits, Oregon's plan- ners developed a method for valuing medical outcomes that stemmed from particular medical interven- tions. This blend of medical fact and value to patients allowed for comparing valuations by introducing cost considerations. Condition-treatment (CT) pairs linked a medical condition with one or more courses of treatment. The goal was to determine the likely incremental medical benefit from a given treatment. In addition, Oregon developed a Quality-of-Well-Being scale to determine the net patient benefit from medical intervention and used a telephone survey to value that net benefit. A cost-benefit ratio was derived, and a prioritization of CT pairs was developed. The article analyzes and evaluates Oregon's use of cost-benefit calculations in the allocation of Medicaid funds, noting that Oregon itself backed away from many of the implications of its cost benefit analysis and that the Americans with Disabilities Act has constrained use of quality-of-life judgments in Medicaid resource allocation decision-making. ~g 1997 Elsevier Science Ltd

Key word~--cost benefit, Oregon, Medicaid, disabilities, rationing, resource allocation

BACKGROUND

Medicaid is a p rogram that provides financing for

medical care to certain categories of indigent

patients. Since its inception, Medicaid has been an

example of cooperat ive federa l i sm-- the federal gov-

e rnment provides the bulk of the funding, but states

must also appropr ia te and spend a substantial

amoun t of their own funds in order to qualify for

*Author for correspondence. t in general, Medicaid does not allow a state to exclude

coverage of specific illnesses; federal law, however, was thought to permit states to exclude organ transplants from their Medicaid program--42 U.S.C. §1396b(i)(1). See Dexter v. Kirschner, 972 F.2d 1113 (9th Cir. 1992); Ellis ex rel. Ellis v. Patterson, 859 F.2d 52 (8th Cir. 1988). Disagreeing with the earlier cases, subsequent court decisions have found that states do not have dis- cretion to exclude coverage under Medicaid for organ transplants. See Pittman v. Secretary, Florida Department of Health and Rehabilitative Services, 998 F.2d 887 ( l l th Cir. 1993); Pereira v. Kozlowski, 996 F.2d 723 (4th Cir. 1993).

the federal matching money (called federal financial participation). The federal government establishes min imum and max imum criteria for eligibility, but within those parameters states have freedom to determine which potentially eligible persons actually qualify for par t ic ipat ion in Medicaid; the federal government mandates tha t certain services be cov- ered under a state 's Medicaid program, but states can choose to add certain opt ional services to the core program required by the federal guidelines. In all cases, the federal government will match qualify- ing state expenditures for all facets of the approved Medicaid program; similarly, states must spend their propor t iona l share on all componen ts of the p rogram if they wish to obta in those federal match- ing funds.

In 1988, Oregon faced a budgetary shortfall for its Medicaid program. The legislature's response was to deny coverage for most organ t ransp lan t s . t The funds saved by that l imitat ion were used to pay for materni ty care for 1500 women. The political/ moral issue was raised in a part icularly po ignant

545

546 James F. Blumstein

way when a seven-year-old boy was denied a bone marrow transplant as a result of the state policy.

In 1989, the Oregon legislature decided to cover all Medicaid eligibles to 100% of poverty; pre- viously, the state had covered those persons up to 67% of poverty. The budgetary trade-off, in order to keep overall program costs in check, was to limit the number of services Medicaid would cover. The basic normative choice that Oregon made at that stage was to cover all potentially eligible persons whose incomes were at or below 100% of poverty. Costs would be held in check by the uniform limi- tation of access to certain services.

To understand the nature of the Oregon plan, it is important to note the choices that were available to Oregon at the time that it made the choice to cover all those eligible to 100% of poverty and to treat all covered Medicaid beneficiaries the same. First, Oregon could specify the funds available and seek bids on an array of services for Medicaid bene- ficiaries. That approach could provide some choices to Medicaid beneficiaries with the market setting the package of services and, implicitly, the prices. Second, Oregon could specify the benefit package and put out that package of services for bid on price. The state was not equipped to specify a ben- efit package at the outset, however. Third, Oregon could develop a prioritization list and price out the component services. The amount of money avail- able would then be a political decision informed by data about the cost of different services and influ- enced by the priorities established through an inde- pendent process of priority-setting.

The Oregon plan reflects adoption of the third approach and very much reflects a physician's per- spective. Under this scenario, providers determine what works, how well it works, and what benefits accrue. The benefits, however, are seen largely in medical terms. The Oregon planners recognized that, even when one thinks strictly in terms of medi- cal outcome, perspectives of patients and physicians may diverge as to what is success and when it occurs. For example, when a physician sees an x- ray of a lung and it is not clear, there will be a con- cern about tuberculosis. For that physician, when the x-ray is clear, success has been achieved. For the patient, however, there may be a recovery period of six to eight months after a clear x-ray. Effectiveness studies tend to measure clinical out- comes; they are not patient-centered in this regard.

*Professional norms and autonomy were respected in the Oregon process; no attempt was made to alter the method of intervention--i.e, to change the production function. The method of diagnosis and treatment was assumed to be a given. Developing alternative, less ex- pensive modes of treatment was not part of the Oregon experiment's agenda. This is in contrast with Tennessee's Medicaid alternative, TennCare, in which change in the production function was a clear goal.

Recognizing this difference in perspectives between physician and patient, Oregon's planners sought to separate fact from value. Clinical consul- tation with physicians would provide data regarding clinical medical outcomes. Judgments about clinical efficacy would result in data regarding the medical consequences of intervention.* These were to be deemed ~facts". But those data would not answer the question of value--how important is the medi- cal outcome from a specified medical intervention on the patient? The challenge undertaken by the Oregon planners was to develop a method for valu- ing medical outcomes that stemmed from particular medical interventions, for blending fact and value, and, ultimately, for comparing those valuations by introduction of cost considerations.

Importantly, and this is a telling point, the values sought to be identified were those of medical out- comes to patients. Patient valuation compared sub- jective valuation of different health states but not a cross-valuation of certain improvements in health status vs other, non-health-related priorities. Reflecting the medical professional model and the physician-based origins of the Oregon prioritization plan, other broader political/symbolic/cultural/econ- omic considerations were not part of the valuation process. At the end of the day, these broader con- siderations were the cause of some considerable ree- valuation and re-prioritization of certain procedures. They remind practitioners and metho- dologists of cost/benefit analysis that it is important to assess accurately the nature of the benefits of a public program, that comparison of those benefits to other publicly provided goods or services is an essential dimension to the evaluation process, and that political/legal/institutional constraints some- times limit the ability of cost/benefit analysts to take into account the full array of costs or benefits.

THE OREGON PLANS AND THEIR METHODOLOGY

In reviewing the Oregon experiment, it is helpful to understand that there have been four plans that have evolved from the planning process.

The first plan/method resulted in the creation of 709 "condition treatment" (CT) pairs, using diag- nostic and procedure codes. A medical condition (e.g. appendicitis) is linked with one or more courses of treatment (e.g. appendectomy). For each CT pair, the planners gathered information on the benefits and costs of the treatment. Planners relied on professional societies to determine what clinical outcomes resulted from specific medical interven- t i o n s - w h a t a patient's health status would be with or without the medical intervention and what prob- ability is attached to the likely outcome. The goal of that process was to determine the likely incre- mental medical benefit from a given course of treat- ment.

The Oregon experiment 547

Beyond the consideration of incremental medical benefit in purely scientific terms, Oregon's planners sought to determine net benefit from the medical interventions. They utilized a Quality-of-Well-Being (QWB) scale* in which the "net benefit" was a comparison of the QWB without medical interven- tion and the QWB with such intervention.t This was an attempt to operationalize in functional terms the consequences of medical intervention, but the concept of "net benefit" itself is incomplete because it does not include an evaluation of those net benefits from medical intervention. Consequently, Oregon's planners had to develop a method for evaluating or weighting the QWB scale.

This evaluation or weighting process was accom- plished by use of a telephone survey of 1001 citizens of Oregon. Each respondent was asked to compare the health state described by the surveyor to death (given a score of zero) or to perfect health (given a score of one). Costs of the services were estimated based on Medicaid records and subjective judgment. A cost-benefit ratio was derived,:~ and a prioritiza- tion of CT pairs was developed and published as an interim proposal. Upon consideration, the result was not accepted by the Health Services

*Oregon used a QWB scale developed by Kaplan and Anderson (1990); it had 24 health or functional states.

tThis analysis can be presented through a formula:

Net Benefit of Service = QWBRx -- QWBNo.

See Eddy ( 1991 ). :~The cost-benefit ratio has been expressed as follows:

Cost-Benefit Ratio

Cost of Service - Net Benefit of Service x Duration'

See Eddy (1991) and Klevit (1991) (p. 914). §For a list of these categories and their description, see

Office of Technology Assessment, 1992 (p. 6). Examples of the categories include: acute fatal (treatment prevents death with full recovery); and acute fatal (treatment prevents death without full recovery). An example of the first category would be appendectomy for treatment of appendicitis; an example of the second type of "acute fatal" would be medical therapy for acute bacterial meningitis,

¶The categories were ranked by the Health Services Commission, which sought to reflect broad community values inferred from a series of public hearings and meetings.

HSee footnote 4 and accompanying text. The Health Services Commission relied on the assessments of treat- ment outcomes furnished by provider groups and by the telephone survey described earlier in text.

**OTA noted a significant difference between the per- ception of Oregon's planners and its own findings regarding the significance of the final line-by-line adjustment. Oregon's planners estimated that 25% of CT pairs were moved in some way during that process. OTA concluded that 53% of CT pairs "moved at least 25 lines from their original positions, and 24 percent of all CT pairs moved up or down at least 100 lines on the list" (Office of Technology Assessment, 1992, pp. 5-7).

Commission, the body appointed by the Governor of Oregon and charged with developing the priori- tized list.

The second approach, which was used to revise the original interim proposal, relied on a three-step process. First, the planners developed a list of 17 service categories and accompanying descriptions.§ Each CT pair was placed in one of these 17 cat- egories. The categories, in turn, were divided into three groups: essential services, very important ser- vices, and services that are valuable to certain indi- viduals. Nine categories were deemed "essential services" and ranked 1-9; four categories were deemed "very important" and ranked 10-13; and four categories were deemed "valuable to certain in- dividuals" and ranked 14-17.¶

Second, once placed within a category each CT pair within that particular category was ranked in accordance with its "net benefit".!l Finally, the Health Services Commission reviewed a preliminary ranked list on a line-by-line basis, using its judg- ment to move selected CT pairs up or down the list. The Commission considered such factors as public health impact, treatment-related cost, and values inferred from public meetings conducted around the state. The cost of funding items on the list was estimated by an actuary, and the legislature determined how many of the CT pairs would be made available to beneficiaries. This plan was even- tually submitted to the federal government for a Medicaid waiver and was turned down by the Bush Administration.

Interestingly, the Office of Technology Assessment (OTA) found that cost was not a major determinant of CT ranking. Despite Oregon's pur- ported use of quantitative data on treatment out- comes and individuals' preferences, OTA concluded that "net benefit" had "surprisingly little effect on the final ranking" of a CT pair (Office of Technology Assessment, 1992, p. 5). What was more important in the ranking process, according to OTA, was the exercise of subjective judgment by the Commissioners in determining the category rankings and, particularly, in making the final line- by-line ranking adjustments.**

Once the Bush Administration turned down its waiver request (Sullivan, 1994), Oregon revised its plan for resubmission to the Clinton Administration, which had taken office. The CT pair list was shortened from 709 to 688. Because of objections from the Bush Administration, Oregon no longer considered quality of life in evaluating the effect of treatment on health outcomes. The results of the telephone survey regarding the pub- lic's health preferences were disregarded. The CT pairs were not placed in broad categories. Instead, Oregon's planners evaluated the relative effective- ness of a treatment primarily on the basis of three "objective" criteria: (1) probability of death; (2) probability of return to an asymptomatic state; and

548 James F. Blumstein

(3) cost of avoiding death. Only professionally sup- plied medical outcomes data were used in this pro- cess, since the introduction of values from the community meetings and the use of the health pre- ference data from the telephone survey were disre- garded. As in the earlier plan/methodology, the subjective judgment of the Commissioners was used as a check on the "objective" ranking.

Upon review, the Clinton Administration approved Oregon's revised submission but only if Oregon eliminated from consideration the "objective" factor regarding the probability of a course of treatment leading to a return to an asymptomatic state. Following the expressions of concern of the Bush Administration, the Clinton Administration concluded that considering return to an asymptomatic state could lead to discrimi- nation against persons with disabilities in viola- tion of the Americans with Disabilities Act (ADA) (Health Care Financing Administration, 1994). As a result, the Oregon plan/methodology actually approved for a Medicaid waiver is a far cry from the original cost-effectiveness proposal. In actual operation, the primary criterion for pri- ority among CT pairs is the ability of the treat- ment to prevent death (which is considered a disability-neutral factor). Average cost of treat- ment can be considered as a tie-breaker factor. Subjective professional judgment by members of the Commission can also be used in setting pri- orities among CT pairs.

THE RATIONALES AND ASSUMPTIONS UNDERLYING OREGON'S APPROACH

The prioritization approach that Oregon embraced must be understood within the historical context from which it evolved. The decision to cut back on funding a certain expensive procedure-- organ transplants--was made in order to provide funds for an expansion of Medicaid coverage. The mindset was that a central policy decision should be made to allocate (or ration) public health care dol- lars in accordance with public priorities. Oregon's policy objective has always been to expand access and eligibility. The basic policy choice in Oregon was to provide Medicaid coverage to those at or below 100% of poverty and not to allocate resources based on eligibility criteria. All those included in the program would be treated uni- formly. Blank check financing for everyone at or below 100% of poverty was understood to be unsustainably expensive. Thus, the plan to prioritize services for Medicaid participants was developed as

*For a discussion of the policy significance of the choice of paradigms--professional vs market-oriented--see Blumstein (1994a).

tFor a discussion of these issues, see Havighurst and Blumstein (1975).

a means for Oregon to live within realistic budget constraints.

Thus, the assumption in Oregon was that the pool of Medicaid funds should be envisioned as a common resource available to all covered citizens on an equal basis. It followed, for the Oregon planners, that this sink-or-swim-together philos- ophy mandated a unitary process. Evidence on preferences was taken and applied in aggregated form. Data on outcomes were analyzed in uniform terms and aggregated. The allocation of Medicaid funds was seen as a matter of public health policy in which a single collective choice had to be made for the expenditure of public resources. And, since the resources were obviously public, government should establish priorities. Reflecting the essential inclusiveness of Oregon's planners, the process should be open, rational, and politically accounta- ble. But the goal for the outcome of the process was quite clearly to provide for a uniform benefit package. The process was seen as a way of defin- ing the scope of public responsibility and individ- ual entitlement.

The Oregon process was very much influenced by a professional paradigm.* A professional consensus was sought on practice protocols. Professional input from professional societies was obtained on appropriate treatments for identified conditions and on the expected effect of such treatments on health outcomes. While reliance on professional judgment in that process is surely reasonable, there is con- siderable evidence that practice patterns diverge widely across small geographic areas. Yet, Oregon did not attempt to influence the existing patterns of practice in order to economize. Existing practice norms were to be ascertained and were deemed to be the benchmarks from which decision-making would begin. Control would remain within the pro- fession and would not be disturbed in an attempt to free up resources for reallocation through reorder- ing practice patterns. The "production function" for the practice of medicine would be undisturbed.

The recognition of the need to value medical ser- vices was a significant insight and advance in tra- ditional professional thinking. All too often, the custom and expectation in the medical community was that resources would be made available to pro- vide all services deemed by the professional commu- nity to be of benefit to patients without the formal consideration of costs. Costs are often taken into account implicitly in the professional decision-mak- ing paradigm; Oregon considered costs and benefits explicitly and thereby challenged some traditional professional thinking.t

There is much to commend in Oregon's approach. There was a consumer/expert partnership not always present in the professional paradigm. The decision-making process regarding prioritiza- tion of public resources was open and accountable. It reflected a responsible grappling with resource al-

The Oregon experiment 549

location issues, including the particularly sensitive one of valuation of benefits. But the Oregon cost- effectiveness approach resulted in the establishment of CT priorities that the political process in Oregon did not accept and that the professional community did not buy into. And the process itself and some of the criteria adopted in the course of the prioriti- zation process apparently ran afoul of the antidis- crimination provisions of the ADA, thereby requiring a thoroughgoing modification of the prioritization process itself. While there were techni- cal problems in the process, I will now focus on the conceptual issues concerning the Oregon prioritiza- tion process.

FRAMEWORK FOR EVALUATING OREGON'S APPROACH

Cost-effectiveness analysis works best when there is (1) a unitary decision-maker (2) allocating a pool of funds (3) in a way that affects all collectively (4) to achieve a defined goal set by a decision-maker. That is, cost-effectiveness analysis is a form of col- lective choice where a decision-maker must decide according to averages or other central tendencies.

Collective choice by public authorities is often necessary. For example, in the realm of public health, measures such as sanitation or spraying against disease-carrying pests must be performed in a collective manner. Decisions regarding the public water supply and its safety must be made in a col- lective manner, and aggregation of benefits is appropriate and necessary. The same is true regard- ing the setting of standards regarding certain en- vironmental health standards such as clean air or water.

While sometimes necessary, however, such collec- tive choice has important disadvantages. Decision- making by aggregation necessarily results in the loss or subordination of individual choices and prefer- ences. When decision-makers rule by averages or central tendencies, there is a loss in sensitivity to the fine-tuning accommodation to individual prefer- ences, and many of those preferences will go unful- filled.

Coping with the problems of collective choice #l the political process

In the political process, we attempt to overcome this disadvantage of collective choice and majority rule in several ways. I will discuss three of those ways.

First, society decentralizes decisions to allow smaller groups to decide on matters that affect them disproportionately (Bermann, 1994; Weiler, 1991). Even if a majority in a broader constituency would decide a matter differently, the decentralized decision-making unit is often empowered to decide

the issue for itself in a way that allows for control by a majority of a smaller group. This permits greater individual influence, as the political trans- action costs are lower in a smaller polity. Further, the system as a whole allows more accommodation to individual preferences because individuals have the ability to choose the political sub-unit in which to participate. This facilitates the choice by individ- uals among values. For example, individuals have choices among states, counties, cities, suburbs, and rural areas. This allows for more freedom to choose from among the competing communal values those in better sync with one's tastes. Thus, in the area of public education, a family can pick a district that more closely accords with its taste in public edu- cation.

Second, society establishes political safeguards for minority rights in certain areas. In some cases, this type of protection is embodied in the Constitution. The American Bill of Rights is an example of this. It provides core freedoms that the government cannot abrogate, even if approved in the political process. Similarly, through statutes such as civil rights laws, society binds itself to respect certain rights of minorities and to confer on those protected citizens legally enforceable claims when laced with discrimination. These consti- tutional and statutory safeguards for individual rights and preferences are examples of the way in which the political system guards against rampant, winner-take-all collective majoritarianism (Federalist No. 9; Federalist No. 10: Guinicr, 1994: Blumstein, 1994b).

A third mechanism used to accommodate individ- ual rights and preferences in the political process is to depoliticize the process. This is the economist's vision. Markets are good at allowing for individual choice. They allow disaggregated preferences to bc accommodated. There is no need for all stake- holders to reach a consensus or a single result Provided that enough consumers are willing and able to purchase a good or service at a price that will produce a competitive return, that good or scr- vice is likely to be supplied in the marketplacc. Even it" those consumers reflect a distinctly minorit? set of tastes, those tastes will be catered to if they are sufficiently widespread to support a provider of those goods or services. No single decision on tastcs is required, and no aggregated one-size-fits-all mode of decision-making is required. This is the rationale behind proposals tbr privatization, use of voucher~ and so forth.

The Oregon approach saw Medicaid as requiring a collective decision by a public entity. The process adopted envisioned a single set of prioritized CT pairs applicable to all Medicaid beneficiaries. The professional model suggested pursuit of a single benefit package for all beneficiaries and identifi- cation of a single best practice for diagnosis and treatment, even though existing evidence on practice

550 James F. Blumstein

variation and clinical uncertainty calls this approach into question (Anderson and Mooney, 1990; Wennberg, 1984).

Pi(falls f rom treating the allocation o f Medicaid .funds as a unitao' decision

A number of pitfalls emerge from this approach to Medicaid. First, private preferences in this area may well not be monolithic; they in fact are likely to diverge. For example, different individuals or groups may have different medical needs. Elderly poor citizens may be able to do without obstetrics services, and they might benefit from participating in plans that forgo offering obstetrics services. That would allow those plans to target their services to a segment of the market and reallocate funds from obstetrics to areas in which the elderly poor may have greater needs such as geriatrics.

The unitary collective choice model of resource allocation assumes the desirability of a monolithic benefit package for all Medicaid beneficiaries in the name of equality. But personal medical services pro- vided to Medicaid patients need not be treated in the same way as setting pollution standards or as establishing safety and health standards in the pub- lic water supply. Those types of public health ser- vices must be consumed collectively; the kind of personal medical services provided through the Medicaid program can be consumed individually. Specialized needs can give rise to niche markets that may allow for the selective targeting of resources. This could allow for the accommodation of individualized private choice and enhanced con- sumer welfare. The existence of differences in need and taste across consumers, when linked with the ability to consume personal medical services indivi- dually and not collectively, suggests that the collec- tive decision-making model of Oregon's Medicaid experiment was not optimal. Oregon's planners may have been operating under a paradigm mistake in this regard, treating Medicaid resource allocation choices as a collectively consumed public health ser-

*At least in theory, Tennessee's Medicaid demonstration (TennCare), see note 16, provides an example of a more disaggregated or decentralized decision-making strategy. Within the funds provided by the capitation payment, each contracting managed care organization (MCO) can modify its benefits package to attract TennCare patients. Tennessee's requirements for cover- age are so comprehensive, however, that such niche marketing cannot as a practical matter under current circumstances allow for significant differentiation in plan design. One could envision within a revised TennCare framework more flexibility in terms of cover- age, so that plans could offer actuarially similar plans that would target younger patients, older patients, per- sons with special needs and so forth. The essential con- trol is that program participants who select one plan would not be able to secure services (exclusive of emer- gency services) not offered by their designated plan.

vice rather than as a set of individually consumed personal medical services of varying priority to different consumers.*

Second, the specification of costs incurred or ben- efits to be derived from the priority-setting process is not altogether clear, and a clearer specification of those costs and benefits in a public, open, and accountable process may not be politically doable or otherwise realistic. Yet, without a clear specifica- tion of costs incurred or benefits to be derived, a rational cost-benefit calculation cannot proceed in a sensible manner. As was the case when Oregon's first interim prioritization list was announced, a plan that cannot clearly identify or define the range of costs to be avoided or benefits sought is likely to run amok.

With respect to costs, there is a clear demoraliza- tion that stems from society's recognizing and accepting overtly that resources for medical care are constrained. And with respect to benefits, much more is at stake than a green-eye-shade evaluation of medical benefits conferred. Programs like Medicaid, which are redistributive in nature, are supported in part because taxpayers are willing to spend money from which they do not derive an im- mediate or direct benefit. These programs appeal to a larger goa l - - to enhance certain general beliefs about the nature of our democratic, compassionate society. The provision of medical services to needy and worthy indigent patients is part of that agenda (President's Commission, 1983). And that agenda includes respecting the cultural norm that life is beyond price. Thus, society's provision of medical care to Medicaid patients involves much more than a technical question of lifesaving or quality-of-life- enhancing. Symbolic values are at issue that go to the heart of our fundamental beliefs about the nature of society and its obligations to its citizens. Oregon's process of public prioritization of medical care for Medicaid beneficiaries therefore implicates much more than determining which conditions war- rant priority and which treatments are best suited to alleviate specified medical conditions (Calabresi and Bobbitt, 1978).

Traditionally, it has been a "basic tenet of our society that we will not give up a life to save dol- lars, even a great many dollars" (Havighurst et al., 1976, p. 140). That life is priceless is a comforting abstraction, a myth that is honored in the breach. But direct head-on confrontation with this "lifesaving imperative" (Havighurst et al., 1976, p. 138) can be "ethically demoralizing" (Havighurst et al., 1976, p. 139), and we often go to great lengths to avoid this type of direct clash with basic beliefs.

The myth of a democratic government, supporting human life at any cost, is a value to which we like to adhere. In general, however, we are unwilling to pay the price necessary to achieve that goal. When a breach of this value is threatened in stark terms,

The Oregon experiment 551

much more may be involved than valuing life. Society's humanitarian self-image may be at stake, and we may be willing to expend considerable sums of money, ostensibly to save a life, but also to pre- serve a valuable myth (Blumstein, 1976).

The crucial insight here is that the structure of the decision-making process "largely determines the degree to which society's deeply held sym- bolic values come into play in particular cases" (Havighurst et al., 1976, p. 142). Thus, the "choice of one procedure rather than another for assigning values to human l i fe . . .may itself have a significant effect on the ultimate value assigned to human life" (Blumstein, 1976, p. 233). And the kind of overt identification of values and resultant prioritization of CT pairs reflected in the Oregon approach is likely to have maximum symbolic effect in this regard. "Formulation of official priorities will likely have a debilitating effect on long-held symbolic values, especially since decisions will often affect clearly identifiable individuals" (Blumstein, 1976, p. 250). Thus, an important societal cost that is difficult to include in the priority-setting decision-making process is the demoralization that derives from the overt breach of the "lifesaving imperative" that inevita- bly flows from an explicit explication of priori- ties.* Some less public form of choice-- preferably by non-government entities such as in- dividuals themselves or managed care organiz- a t i ons -wou ld reduce this set of concerns, although concededly it does raise legitimate ques- tions about how and how fairly those decisions are being made (Katz and Capron, 1975. pp. 178-79. 189 92. 217).

THE USEFULNESS OF COST-EFFECTIVENESS ANALYSIS 1N THIS ARENA

Oregon's attempt at using cost-effectiveness analysis did not achieve the objectives sought by its promoters. First, the original methodology led to a set of priorities that did not gain political or pro- fessional acceptance. Second, by including the qual- ity-of-life elements so important to the entire schema, the Oregon approach apparently ran afoul of the antidiscrimination provisions of the ADA, This section will consider these two issues.

*For recognition of the importance of this consideration in explaining why Oregon's first interim prioritization list was politically unacceptable, see Hadorn (1991).

fSee Hadorn (1991) (describing the lifesaving impera- tive as the Rule of Rescue).

:~"The more visible and more directly accountable the government becomes, the more nearly analogous to capital punishment decisions these issues become. The possible result would be a subtle form of institutional blackmail, inducing society to spend more resources in this area than it might otherwise choose" (Blumstein, 1976, p. 252).

Can cost-effectiveness analysis succeed in allocating medical resources?

Critics of the original Oregon approach have asserted that the cost-effectiveness methodology is inherently flawed in the allocation of Medicaid resources because it does not take into account the broader symbolic issues associated with the lifesav- ing imperative.f That is, "cost-effectiveness analysis is inherently incapable of capturing the value people place on lifesaving technologies and rescuing indi- vidual, identifiable lives" (Eddy, 1991, p. 2138). In response to this criticism, David Eddy has argued that the Oregon methodology was flawed because of technical errors of implementation (Eddy, 1991. p. 2138). He asserts that failure to account properly for the lifesaving imperative "was due not to any inherent features of cost-effectiveness analysis but to . . . the measure of benefit used" (Eddy, 1991, p. 2138). Eddy contends that, with the benefit defined properly, cost-effectiveness analysis can be used appropriately to "maximize the total amount of goodness that can be produced with the available resources" (Eddy, 1991, p. 2140). He views the pro- blems confronted in Oregon as those of technical implementation not those of an inherent limitation to the methodology.

There are several points that are worth making here. First, Eddy may be correct that "'cost-effec- tiveness analysis is philosophically neutral" (Eddy, 1991, p. 2141), but that is quite different from say- ing that that methodology in this particular context is outcome-neutral. As discussed earlier, the approach that is used in the context of valuing life or health is likely to influence the outcome. "'There may be important results that stem from the choice of one procedure rather than another for assigning values to human life" (Blumstein, 1976, p. 233). And, in the allocation of medical resources, an overt public process that assigns priorities to certain health states is more likely to be susceptible to the kind of "symbolic blackmail" (Blumstein, 1988, p. 865) discussed above. The more visible and overt the process of resource allocation, and the more closely identified with government, the more that the lifesaving imperative comes into play. The pro- blem does not necessarily lie with the methodology of cost-effectiveness as applied in this context. Rather, because the express articulation of prin- ciples of allocation unnecessarily' and provocatively is inevitably destructive of importantly held sym- bolic values, cost-effectiveness analys is-even if applied with maximal sophistication--may be ill- suited to this purpose.+ +

Therefore, both critics and supporters of cost- effectiveness analysis in the allocation of Medicaid resources may not have addressed the most compel- ling concern about Oregon's use of cost-effective- ness analysis. The professionally oriented explicit prioritization of CT pairs exacerbates the symbo-

552 James F. Blumstein

lism associated with the lifesaving imperative. The goal should be to defuse not accentuate the symbo- lism associated with medical care decision-making. This would keep the government's decisions at the macro level regarding resource allocation, and it would delegate to others the micro decisions of how the gross dollars allocated to Medicaid should be apportioned.

Interestingly, Oregon never attempted to use cost-effectiveness analysis for making the macro-al- location decision regarding the total number of dol- lars to be appropriated to the Medicaid program. All along, that macro-allocation decision was recog- nized as a political choice by the governor and the state legislature. Yet, the decision on macro-funding is very much at the statistical level and can be pur- sued with minimal influence by the lifesaving im- perative. Identifiable persons are not yet involved, and appropriate trade-offs between needs of Medicaid beneficiaries and others can be discussed somewhat more dispassionately at that stage of the resource allocation process. If one were to design a cost-effectiveness methodology at that level in the process, it would likely have less of the symbolic effect than that experienced by Oregon's planners, who focused at the micro-allocation level.

Perhaps a better approach would be for the state legislature to determine what level of resources it chooses to allocate to Medicaid. Once that macro- allocation choice is in place, the state would allow bidders to present actuarially equal packages to in- dividual Medicaid beneficiaries who could select their plan. This would place the package design and implementation issues--the micro-allocation choices that are so fraught with symbolic issues associated with the lifesaving imperative--outside the ken of government. The trade-offs would be made in the

*The TennCare program, Tennessee's Medicaid alterna- tive, adopts some of these ideas. See note 12. Tennessee set out an amount of money per beneficiary. It contracted with managed care organizations, which were responsible for plan design and implementation. The conceptual problem for TennCare is that the state specified the benefit package as well as specifying the capitation rate per beneficiary. Without the kind of homework done in Oregon, Tennessee's planners were unable to state with confidence whether, even with expected changes in the ++production function" of medical care delivery, the capitation rate was adequate to cover mandated services. Over time, Tennessee may be able to place plan design in the private sector, leav- ing to the market the development of benefit packages and leaving to the state the setting of the capitation rate with appropriate monitoring of quality of care on behalf of Medicaid patients.

tThe federal government expressed concern that people without disabilities would undervalue the quality of life of persons with disabilities. Robert M. Kaplan (1995) has strongly objected to this component of the federal government's analysis, concluding that "overall prefer- ences for health states appear to be quite similar" (p. 52) for the disabled and non-disabled.

private sector, and allowance for different benefit packages would accommodate the desirability of allowing niche markets to be targeted and served most efficiently.* Cost-effectiveness analysis, in this scenario, could be useful for managed care organiz- ations in designing their packages of service. But collective valuation of services by the government ++outside" the political process would be avoided, except as possibly implemented at the macro-allo- cation level.

Implications of the Americans with Disabilities Act

When in 1992 the Bush Administration turned down Oregon's request for a Medicaid waiver, it relied on concerns about possible violation of the antidiscrimination provisions of the Americans with Disabilities Act. This legal landmine had apparently eluded or not been taken seriously by legal consult- ants who had analyzed these issues during the run- up to the waiver review process.

In its decision, the Bush Administration left open the precise basis for its rejection of the Oregon pro- posal. Two theories were advanced. One expressed concern about bias in the telephone survey. Since respondents were being asked to evaluate various health states, the government seemed worried that the telephone sample did not adequately reflect the views of persons with disabilities, who might have different views from non-disabled persons about the utility of life in a disabled state.t That concern led Oregon to disregard the results from its telephone survey, thereby eliminating any consideration of uti- lity in its revised, resubmitted proposal. While such a concern about bias in the telephone sample may or may not be correct, the answer to that question would not imperil a state following in Oregon's footsteps from employing the Oregon methodology. Since Oregon arguably did use a biased survey and was unable to correct that alleged error in time for its resubmission, it had no practical alternative (other than to litigate and suffer even greater delay) but to disregard the results of that survey in devel- oping its revised prioritized list. But the dispute about sample bias does not imperil the Oregon methodology under the ADA; it does require some consensus about whether the federal government is correct in its concern about such bias, But once that technical dispute is resolved, the approach would not be a problem if this minimalist objection to Oregon's plan under the ADA were the sole diffi- culty.

Unfortunately, the Bush Administration also expressed a more maximalist potential objection to Oregon's methodology under the ADA. This maxi- malist concern questioned the inclusion of quality- of-life and functional health status judgments in Oregon's prioritization process (Sullivan, 1994). Clearly, if Oregon were unable to use such indi- cators, an essential component of the Oregon plan and methodology would be barred by the ADA.

The Oregon experiment 553

In its waiver resubmission, Oregon excluded from consideration any utility preferences derived from its telephone survey. Ostensibly, Oregon concluded that the minimalist objection under the A D A pre- vailed; accordingly, it agreed to disregard data from the allegedly biased sample. In effect, Oregon no longer had a basis for prioritization based on the valuation of health states. However, Oregon did include as a criterion of prioritization consideration of the probability of moving from death or from a symptomatic state to an asymptomatic state. In granting approval to the revised waiver request, the federal government prohibited Oregon from using either of these factors. The federal government apparently adopted the maximalist view of the ADA, concluding that expressing preference for asymptomatic states could result in devaluing health states with residual disabling symptoms (Health Care Financing Administration, 1994). As a result of this federal determination, Oregon was unable to prefer interventions that were more likely to result in a return to an asymptomatic state to interven- tions that would leave a patient with residual symp- tomatic conditions.

Oregon revised its prioritization list in light of the government 's decision and limited its prioritiza- tion process to consideration of probability of pre- venting death. Prevention of death was deemed to have no A D A implications, since that was a status- neutral factor. If any CT pairs were tied on that basis, then the average cost of treatment would be considered as a tie-breaker, with the higher cost CT pair being ranked lower. If any CT pairs were tied on both those elements, they were ordered alphabe- tically by diagnosis. Subjective judgment of the Commissioners could shift CT pairs up or down the list. That is what was left of the prioritization pro- cess when the program was implemented with a fed- eral waiver in hand.

Quite clearly, the federal government 's maximal- ist interpretation of the A D A leaves little room for the kind of prioritization and cost-effectiveness pro- cess contemplated by Oregon. In practice, no con- sideration of utility preferences is part of the implementation plan. And quality-of-life consider- ations seem prohibited by the federal government 's interpretation of the A D A . Under the circum- stances, the prognosis for use of an Oregon-style cost-effectiveness methodology seems bleak indeed.*

REFERENCES

Anderson, T. F. and Mooney, G. (1990) Medical practice variations: where are we? In The Challenges o[' Medieal Practice Variations, eds T. F. Anderson and G. Mooney. Macmillan, Basingstoke.

*For a further consideration of these discrimination issues, see Mehlman et al. (undated), Crossley (1995) and Peters (1995).

Bermann, G. A. (1994) Taking subsidiarity seriously: fed- eralism in the European Community and the United States. Columbia Law Review 94, 331~,56.

Blumstein, J. F. (1976) Constitutional perspectives on governmental decisions affecting human life and health. Law and Contemporary Problems 40, 231 305.

Blumstein, J. F. (1988) Thinking about government's role in medical care. St. Louis Universi O, Law Journal 32, 853 867.

Blumstein, J. F. (1994a) Health care reform and compet- ing visions of medical care: antitrust and state provider cooperation legislation. Cornel/ Law Review 79, 1459 1506.

Blumstein, J. F. (1994b) Federalism and civil rights: comp- lementary and competing paradigms. Vanderhilt Law Review 47, 1251 1301.

Calabresi. G. and Bobbitt, P. (1978) Tragic Choices. Norton, New York.

Crossley, M. A. (1995) As we lay dying: medical futility and disability discrimination. Iowa Law Review 81, 179 259.

Eddy, D. M. (1991) Oregon's methods: did cost-effective- ness fail?. Journal 0/" the American Medical Associathm 266, 2t35 2141.

Federalist No. 9 (Hamilton) (1961) In The Federalist Papers. ed. C. Rossiter. New American Library, New York.

Federalist No. 10 (Madison) (1961) In The Federalist Papers, ed. C. Rossiter. New American Library, New York.

Guinier, L. (1994) The Tyranny o[' the Mq]ority: Fundamental Fairness in Representative Democraey. Free Press, New York.

Hadorn, D. C. (1991) Setting health care priorities in Oregon: cost-effectiveness meets the Rule of Rescue. Journal 0[" the American Medical Association 265, 2218 2225.

Havighurst, C. C. and Blumstein, J. F. (1975) Coping with quality/cost trade-otis in medical care: the role of" PSROs. Northwestern University Law Review 70, 6 68.

Havighurst, C. C., Blumstein, J. F. and Bovbjerg, R. (1976) Strategies in underwriting the costs of cata- strophic disease. Law and Contemporary Probh,ms 40, 122 195.

Health Care Financing Administration(1994) Special terms and conditions regarding the Oregon plan. L~'sues hi Law atut Mediehw 9, 418 422.

Kaplan, R. M. (1995) Utility assessment for estimating quality-adjusted life years. In Vcth~bzg Health Care: Costs. Bemf[its, and Effk'etiveness ~[' Pharmaceuticals and Other Medical Technologies, ed. F. A. Sloan, pp. 31-60. Cambridge University Press, Cambridge, New York.

Kaplan, R. M. and Anderson, J. P. (1990) The General Health Policy Model: an integrated approach. In Quali O, ~?/ Li[~" Assessments in Clinical Trials. ed. B. Spilker, pp. 131 149. Raven, New York.

Katz, J. and Capron. A. M. (1975) (~aastrophic Diseases. Sage, New York.

Klevit, H. D. (1991) Prioritization of health care services: a progress report by the Oregon Health Services Commission. Archives ~?/' btternal Medicine 151, 912 916.

Mehlman, M. J., Durchslag. M. R. and Neuhauser, D. (undated) W'hen Do Health Care Decisions Discriminate Against Persons with Disabilities? [typescript available from authors at Case-Western Reserve University School of Law, Cleveland, OH].

Office of Technology Assessment (OTA) (1992) Evaluation of the Oregon Medicaid Proposal (Summary). OTA, Washington.

554 James F. Blumstein

Peters, P. Jr (1995) Health care rationing and disability rights. Indiana Law Journal 70, 491-547.

President's Commission for the Study of Ethical Problems in Medicine and Biomedical and Behavioral Research (1983) Securing Access to Health Care: A Report on the Ethical Implications of Differences in the Availability of Health Services. President's Commission, Washington, DC.

Sullivan, L. W. (1994) Letter from Secretary, United States Department of Health and Human Services, to Barbara Roberts, Governor of Oregon (3 August 1992), reprinted in Issues in Law and Medicine 9, 409.

Weiler, J. H. H. (1991) The transformation of Europe. Yale Law Journal 100, 2403-2483.

Wennberg, J. E. (1984) Dealing with medical practice vari- ations: a proposal for action. Health Affairs 3, 6-32.