The Inescapable Complexity of Decision making: Ethics, Costs, and Informed Choices
Be Prepared for Sudden Aggravation.
Construction sign, Maryland Department of Transportation
The strategies discussed in Chapters 4 and 5 can encourage more consistent clinical practice, reduce the occurrence and costs of inappropriate care, and increase the provision of appropriate care. They can thus help improve the value received for expenditures on health care. In themselves, however, these strategies do not necessarily lead to commonly held conclusions about how to distribute, organize, or pay for health care. Decisions in these arenas are thoroughly entangled with debates about reforms in health care financing and delivery.
During its deliberations, the committee became most engaged in debate over health care reform when it considered how guidelines for clinical practice might be used as tools to constrain costs. The committee's consideration of this issue did not—and was not intended to—lead to proposals for financing and delivery system reform. Such proposals would have exceeded the committee's charge. However, the final chapter of this report concludes with a few comments on the relationship between guidelines and some proposed directions for reform.
The first section of this chapter highlights some central ethical questions related to the development and use of practice guidelines in making everyday medical decisions and in adopting social policies that may benefit some people at the expense of others. The discussion then turns to health care costs and the responsibilities of guidelines developers to consider these costs in their recommendations. The last sections of the chapter consider the issue of informed consent and the concept of basic or minimum care. Many controversial issues are raised in this chapter. The intent is to present
background information, examples, and different sides of debates on these issues clearly and fairly so that the reader can consider them and perhaps say "I disagree."
The discussion here takes as a starting point the propositions offered in Chapter 1 about how developers of guidelines can improve the knowledge base for making day-to-day clinical decisions and for formulating policies to affect the cost, quality, and availability of health care. To reiterate, every set of guidelines should be accompanied by statements about the strength of the evidence behind the guidelines and by projections of the relevant health and cost outcomes. In building the case for or against particular courses of care, scientific evidence takes precedence over expert subjective judgments. Guidelines should be accompanied by documents that disclose the procedures followed in the development process, the participants involved, the evidence used, the rationales and bases for decisions, and the analytic methods.
More often than not, developers of guidelines will find that particular clinical interventions will not be backed by a clear-cut scientific case. Another, perhaps more difficult challenge will be to distinguish facts from values; self-interest and other biases may be hard to discern even for parties making every effort to be objective. These problems, which are hardly unique to practice guidelines, should be acknowledged candidly and tackled with special sensitivity to their ethical implications.
As guideline developers move in the directions outlined here, they will describe how compelling is the case for the use of specific services in specific circumstances. They will thereby inform but not necessarily dictate answers to ethical or policy questions such as where to draw lines between care that is covered by insurance plans and care that is not.
AN ETHICAL CONTEXT
Members of the IOM committee brought to the study several ethical concerns about the development, use, and evaluation of clinical practice guidelines. These concerns were echoed and amplified in the study's site visits, public hearing, focus groups, and other activities and in a paper (Povar, 1991) commissioned by the committee to help it explore the ethical sensitivities and complexities associated with practice guidelines. A pervasive theme was the real and potential conflicts between individual and collective views about ethical obligations and standards in health care.
Ethical Obligations of Individuals
In discussions about health care, most ethical analysis focuses on the practitioner's ethical obligations to individual patients. These obligations
are generally viewed as doing good, avoiding harm, respecting patient autonomy, and treating patients equitably (Beauchamp and Childress, 1983; President's Commission for the Study of Ethical Problems in Medicine and Biomedical and Behavioral Research [hereafter, President's Commission], 1983; Jonsen and Toulmin, 1988; McCullough, 1988; Brennan, 1991a; Povar, 1991)1 (Professional obligations in the form of honesty, competence, and avoidance of conflict of interest are assumed here.)
Physicians and other practitioners may face conflicts among ethical obligations. A professional's obligations to do good and avoid harm may conflict with the obligation to respect patient autonomy. For instance, competent patients may make choices that appear irrational to physicians but that conform with the patients' own values and priorities (Brock and Wartman, 1990). A case in point is the patient with end-stage renal disease who prefers greater control and freedom in his or her life and so declines to follow difficult dietary, drug, dialysis, and other regimens despite the understood medical risk (IOM, 1991d). When the physician doubts that the various risks and benefits are or can be understood by the patient, he or she may feel even more troubled by conflicting ethical duties.
During the past two decades, patient autonomy has been increasingly emphasized and the paternalistic substitution of professional for patient judgment correspondingly criticized (President's Commission, 1983; Kapp, 1989; Povar, 1991). Single-minded emphasis on autonomy, however, has also been challenged—for three reasons. First, it seems to imply that physicians need only be technicians without being committed to their patients' best interests. Second, it appears to suggest that patients do not need the counsel, judgment, and assistance of physicians and other professionals (McCullough, 1988). Third, it assumes too strongly that patients can always understand the information and the options being presented well enough to make informed decisions. This third issue is not restricted to debates about pa-
tient autonomy and paternalism; it surfaces more directly in discussions of informed consent and in proposals for health care reform based on consumer choice among competing health plans. A different issue is what patient autonomy implies with respect to an individual's assuming financial or other responsibility for the consequences of noncompliance with recommended health practices.
Ideally, practice guidelines should strengthen the dialogue between patient and physician. They should serve the objective of patient autonomy by being as clear as possible about the evidence and rationale for guideline recommendations, the outcomes expected for alternative courses of care, and the ways patients may view these outcomes.2 For example, guidelines that carefully present evidence about how test results will or will not affect patient management or patient outcomes can help physicians distinguish between a mere ''quest for diagnostic certainty" (Kassirer, 1989) and a quest for information that makes a difference in decision making and potential outcomes. In general, guidelines for clinicians and materials developed specifically for patients should be designed to help physicians and patients discuss recommendations more fully in terms relevant to patients; they would thereby demonstrate—rather than merely imply—the "connection between [the physician's] clinical judgment and the best interests of patients" (McCullough, 1988, p. 461).
Ethical Obligations of Collective Social Systems
At the organizational or societal level, ethical analysis becomes particularly complex as it confronts what it means to do good, avoid harm, respect autonomy, and act fairly. Doing good or avoiding harm for a specific individual may sometimes conflict with doing good or avoiding harm for patients (or potential patients) collectively.
For example, when society requires immunizations, reporting of communicable diseases, and similar measures, collective interests in disease prevention generally override personal interests in privacy and autonomy. Less straightforward and less widely accepted are principles regarding services for terminally ill individuals, especially patients with formal "Do Not Resuscitate" orders (Lo, 1991). The individual's (or family's) interest in self-determination may, on the one hand, conflict with traditional professional and social interests in preserving life; on the other hand, the patient's interest may confront institutional interests in limiting resources for care that is seen as nonbeneficial. Other conflicts are possible as well. In the hospice and nursing home case presented in Chapter 3, the regulations re-
garding patient restraints were intended to protect individual patient dignity; the institution, however, must also consider the safety of other patients and staff.
Much of the debate over divergent interests revolves around issues of cost-effectiveness and rationing or allocation of limited resources among alternative uses (Weinstein and Stason, 1977; President's Commission, 1983; Pellegrino, 1986; Callahan, 1987).3 Such debate often highlights trade-offs between providing very expensive services (such as transplants) to a few individuals and providing less expensive services (such as prenatal care) to many more individuals (Redelmeier and Tversky, 1990; Egan, 1991; Fox and Leichter, 1991). The rationing debate, however, should not focus on expensive versus inexpensive health care services per se. The issue is not the expense per unit of service but, rather, the expense per unit of benefit (for example, years of life or freedom from pain). High-volume services with low unit costs are less dramatic but not necessarily less important than very expensive, low-volume services.
Systems must inevitably make trade-offs among alternative ways of using available resources to benefit large groups (their members). 4 Guidelines, in the form of standards of minimum, necessary, or basic services to be covered by public or private health benefit plans, have been suggested as one vehicle for determining these social allocations (Hadorn, 1991a,b,c). A system that has attempted to use guidelines to inform decisions about resource allocation is the Group Health Cooperative of Puget Sound (GHCPS). GHCPS has developed risk-based guidelines for various preventive services. Its process for developing these guidelines considers the literature on the services, other guidelines, and the organizational objectives and capabilities of GHCPS. A major objective has been to increase the proportion of high-risk individuals who undergo recommended screening. The GHCPS
screening mammography guidelines, adopted in April 1988, recommend yearly screening for women aged 40 to 49 with previous breast cancer or abnormal breast tissue or with two or more first-degree relatives with breast cancer; biennial screening for women with one first-degree relative with breast cancer; triennial screening for women with at least one minor factor; and no screening for other women in this age group except on physician referral. (This guideline was reproduced in IOM, 1990c.)
GHCPS moved from guidelines development to guidelines implementation when it sent 67,000 female members over the age of 40 a risk assessment questionnaire and invited women to come for screening as indicated by their level of risk (Thompson et al., 1988). A computerized information system recorded the responses and incorporated them in the patient's medical record; the system now provides monthly reports to physicians on the status of the women in their practices. The case-finding rate in this program is considerably higher than the rate reported in non-risk-based programs.
Screening guidelines are to some degree atypical examples of existing practice guidelines. They are commonly accompanied by information about costs, and recommendations often reflect explicit or implicit judgments of cost-effectiveness. That is, social as well as individual benefit is considered, albeit in the context of choices among health services rather than between health and nonhealth services.
Another IOM committee (1990i) has argued that quality care should be evaluated on the same scale for rich and poor systems, rural and urban settings, academic and nonacademic institutions. That panel recognized that resources will affect decisions and actions. It went on to argue, therefore, that quality assurance and improvement programs should be able to identify (1) how and to what degree resource constraints do in fact affect the structures, processes, and outcomes of health care; and (2) which agent(s) are responsible for such constraints (Lohr and Harris-Wehling, 1991). Thus, although judgments about quality are distinct from judgments about appropriate uses of resources, processes of quality assessment and improvement cannot ignore how the latter affects the former.
GUIDELINES, COSTS, AND DECISIONS
The issues raised above involve two somewhat different meanings of value. The first meaning conveys the normative aspect of the word: what should be the responsibilities of individuals and governments, and how should limited resources be distributed? The second meaning has a more empirical slant: given agreement on basic definitions and assumptions, what is the net benefit and cost-effectiveness of a particular intervention? Information about value in the second sense contributes to, but does not dictate, decisions about the pursuit of values in the first sense.
Why Present Information About Costs?
This report recommends that every set of clinical practice guidelines include information on the cost implications of alternative preventive, diagnostic, and management strategies for the clinical situation in question. The rationale is that this information can help potential users, who must take financial and other resources into account, to better evaluate the potential consequences of different practices. The reality is that this recommendation poses major methodological and practical challenges (Weinstein and Stason, 1977; IOM, 1985; Russell, 1986; Detsky and Naglie, 1990; Eddy, 19901, 1991c,d).5
With respect to the rationale for cost estimates, their purpose is to inform decision making by relating the expected costs of care to the outcomes expected from that care—that is, by projecting the cost-effectiveness or value of the services in question. To be relevant to decision making that is, to the making of choices among alternatives—cost estimates should cover not just a single option for care but also the major (that is, reasonable) alternative or alternatives. The alternative(s) could be watchful waiting, doing nothing, or a different kind of intervention (e.g., a different drug or medical rather than surgical treatment).
In making cost projections, the estimates should go beyond the immediate costs of managing a clinical problem or completing a procedure to encompass the costs related to follow-up care, supportive services, and other steps necessary for the service to make a difference to life expectancy, functional status, or some other result that matters to the patient. For example, the so-called downstream costs of treatment for individuals whose cancers are discovered by screening must be included in the cost of a screening policy, because treatment is essential to make a difference to health. Likewise, cost estimates should include the costs of further testing to pursue false-positive results. The costs associated with not providing the screening or other intervention in question need to be estimated so that net costs or savings can be identified. Where estimates are subject to substantial uncertainty, analysts can use different estimates to indicate how sensitive the projections are to different assumptions. (For a discussion of the practical decisions and tradeoffs involved in making projections, see Rettig, 1991a.)
Cost estimates are most helpful if they show separately the cost of each major component of care. For example, estimates for a screening program might show the cost of initial screening, the cost of follow-up for patients
with positive results in the initial screening, and the cost of treatment necessary to affect health outcomes in a clinically meaningful way. Detail such as this helps those who may use the guidelines to understand better the resource implications of different choices. It also focuses attention on areas in which future improvements in the process can make the most difference to costs.
Clearly, cost-effectiveness analysis and estimation must involve both clinicians and experts in cost-effectiveness analysis. Preferably, some individuals who have both kinds of expertise can be involved. As more clinicians are trained in techniques needed for guidelines development, this cadre of individuals should grow.
The committee discussed extensively what might be expected of cost-effectiveness analysts in the process of guidelines development. Some committee members took the position that analysts should be involved from the early stages of guidelines development. Such experts might, for example, point out that the panel should provide analysts with assumptions about how a patient will be treated once a problem is discovered through screening. This approach suggests that guideline development work should be done "all at once," by the same group of people. What is closer to reality is a "partitioned" approach that deals with cost-effectiveness analyses and judgments in stages. The drawbacks here are that partitioning increases communication costs (between panel members and the analyst) and heightens the chances that the analyst will misinterpret the panel. The compromise may be to encourage guidelines panels to work systematically toward incorporation of cost-effectiveness analysis into their processes.
Again, ideally, those estimating costs for specific health care services or procedures would examine costs and health outcomes with the five questions below in mind.
What evidence suggests that the services are likely to affect outcomes for the condition or intervention being considered?
What groups at risk are most likely to experience benefits or harms from the proposed course of care and its side effects?
What is known about the effects of different frequencies, duration, dosages, or other variations in the intensity of the intervention?
What options in the ways services are organized and provided (for example, size of institution, type of personnel used, experience of personnel, volume of service provided) can affect the benefits, harms, and costs of the services?
What benefits, harms, and costs can be expected from alternative diagnostic or treatment paths, including watchful waiting or no intervention?
Unfortunately, for the vast majority of treatment decisions, developers of guidelines will find that answers to these questions are in short supply. There are several reasons why this is true. First, scientific evidence about benefits and harms is itself incomplete, as has been noted elsewhere in this report. Second, basic, accurate cost data are scarce for the great majority of clinical conditions and services. Third, data on charges may be available, but many significant analytic steps and assumptions are typically required to treat charge data as cost data. Fourth, techniques for analyzing and projecting costs and cost-effectiveness are complex, evolving, and not readily applied by novices. Fifth, and most significant for this discussion, developers of guidelines have, for the most part, not considered costs as a relevant, ethical, or practical subject for their deliberations. In third and fourth areas, further research and development to improve techniques for cost-effectiveness analysis is an important need.
Some guidelines do provide information on costs, but this practice is not yet common. Even the guidelines in Common Diagnostic Tests (Sox, 1987, 1990) and Common Screening Tests (Eddy, 1991a), compilations that stress in their prefaces and elsewhere concerns about excessive costs, do not uniformly provide information about the cost-effectiveness of different tests used under different circumstances. Likewise, the U.S. Preventive Services Task Force discusses the cost implications of some but not all of the services covered in its 1989 report. Still, the analyses presented in the first two volumes cited here and under way at AHCPR and elsewhere should serve as models for other groups.
Should Developers of Guidelines Go Further?
Explicit Judgments about Cost-Effectiveness
Providing information about costs does not guarantee that such information will be used—even now, after years of growing desperation about the escalation of health care costs. This gap between availability of information and action on that information led the committee to consider a recommendation that guideline developers include cost-effectiveness as an explicit criterion for judging or recommending what constitutes appropriate care. Judgments of this kind are made to some degree now, but the role that costs play in such judgments may not always be clearly described in guidelines or related materials.
After much debate, and with some vigorous dissent, the committee concluded that initial developers of clinical practice guidelines need not use economic or cost criteria as explicit bases for recommendations on what constitutes appropriate care for particular clinical problems. Put different-
ly, although guidelines should be accompanied by projections of health outcomes and costs, the specific recommendations for clinical practice can stand on sound assessments of clinical evidence and carefully derived expert judgment.
The committee is not saying that judgments of cost-effectiveness should be or can be avoided. Governments, health benefit plans, health care providers, and others must make such judgments, although they may not always do so explicitly and rationally. The committee also is not saying that developers of practice guidelines should never make such judgments. In particular, when those developing or cooperating in the development of guidelines are also the intended users, judgments of cost-effectiveness may be sensibly integrated into the process. Thus, HMOs, hospitals, and others may weigh costs against expected benefits in making judgments about drug formularies, equipment purchases, testing protocols, and other matters.
The committee decided not to insist that guideline developers employ cost-effectiveness as a decision making criterion for two reasons. First, committee members could not agree that guidelines developers were, from a policy perspective, the right source of authoritative judgments about cost-effectiveness, and several feared that such judgments would complicate the resource decisions of government policy makers, health plan managers, and others. Second, committee members could not agree that their recommendations should go beyond the demanding standards for guidelines that they had already formulated.
Given the present state of guidelines development, adding information about costs will be both a major contribution and a major challenge. Developers of guidelines are, for the most part, still struggling with relatively meager financial resources, scarce data, limited methodologic capacity, unpredictable political support, professional hostility, and a short and largely unevaluated record of performance. Even the presentation of cost information has the potential to undermine the clinical judgments presented in guidelines if potential users think cost considerations are driving the recommendations about clinical care. Still, the committee believes that the effort is important if guidelines (and the discussion accompanying them) are to be useful to decision makers.
In any case, whether developers of practice guidelines only provide cost estimates or choose also to make recommendations based on cost-effectiveness considerations, they must involve individuals with relevant expertise in cost-effectiveness analysis and cost projection in the development process. Further, they should disclose the role that cost information played in their judgments so potential users can assess the extent to which that information drove specific recommendations. Finally, developers of guidelines should, to the extent practicable, specify the settings, payers, and patient groups for which the guideline is being developed or for which it is appropriate.
A Modest Proposal
Even when guideline developers choose not to employ cost-effectiveness as a criterion in formulating statements about appropriate care, they may still be able to state recommendations in ways that help practitioners, patients, and policy makers reach decisions in the face of constraints on individual or system resources. Specifically, they can clearly identify how compelling is the case for particular services or courses of care under particular clinical circumstances. The strength of the scientific evidence (and, secondarily, the strength of the expert consensus) and the nature and importance of the projected health benefits and harms are the central elements in this process.
Depending on available scientific evidence and expert consensus about alternative courses of care, developers of guidelines have several options in formulating statements about appropriate care. Some of these scenarios may be relatively theoretical and rare, but they highlight the ways in which recommendations may relate resources and outcomes.
In triage situations (such as battlefields or emergency departments overwhelmed by local disaster) in which personnel, space, time, and other resource limits are critical, fixed, and immediate, guidelines exist to help practitioners determine when to provide or to withhold care based on comparisons of expected net benefit to individual patients. This is a generally understood point but worth restating in this context.
If the existence or importance of benefits and harms of a familiar service or technology is unclear, guidelines might state that practitioners can forego that particular intervention and still be considered professionally prudent. This gives decision makers more leeway to consider cost factors. (In today's economic climate, new technologies may be held to a stricter standard that requires more definitive information and arguments about benefits, harms, and costs.)
If evidence is sufficient to support several treatment options that have similar costs but different mixes of risks and benefits, respect for patient preferences generally would warrant informed patient choice. If costs of these treatment options differ, guidelines can illuminate but not answer the question of whether patients or third parties should be responsible for the costs of the more expensive option. They certainly cannot dictate what a patient's preference should be for different combinations of risks and benefits.
If evidence indicates that alternative courses of care differ greatly in cost but produce health outcomes and side effects that are similar clinically and that are experienced similarly by patients, then it is reasonable (some would say ethically required) for guidelines to make a judgment based on
cost-effectiveness and recommend explicitly that the less costly alternative be routinely used. A case in point might involve appropriate antimicrobial agents for common upper respiratory infections. Newer-generation antibiotics (or broad-spectrum agents) are likely to be more costly than older (or narrower spectrum) products. When no marginal therapeutic benefit is to be expected from newer or more complex agents, guideline developers may quite reasonably recommend the less expensive agents.
Even when developers of guidelines do not factor costs directly into their recommendations, they or their sponsoring organizations might still regard the publication of guidelines as an opportunity to present opinions or recommendations based on costs. For example, when the Annals of Internal Medicine and other professional journals publish a set of guidelines, an editorial or guest commentary could consider how providers and financers, taking resource constraints into account, might act on the guidelines.
The next sections of this chapter approach the question of what is worth doing or recommending from somewhat different perspectives. The first involves the patient as decision maker and the conditions for informed consent. The second relates to the physician as decision maker. Each discussion attempts to suggest how the "compelling case" approach outlined in the introduction to this chapter may be helpful.
THE PATIENT AS DECISION MAKER: WHAT IS INFORMED CHOICE?
Good medical care requires that decision making be shared, to varying degrees, between practitioners and patients. This message comes from (1) accumulating research on outcomes and effectiveness of health care, (2) case law on the issue of informed consent, and (3) the consumer movement of the past 30 or so years. In supporting shared decision making, guidelines may serve as a basis for physician communication with patients or as a starting point for informational materials prepared specifically for patients and consumers (or their families or other representatives). The following discussion looks first at questions typically raised under the rubric of "informed consent" and then turns to some issues often considered under the heading of "patient preferences."
Patient communication and information can serve at least three objectives. One is to help patients choose among possible strategies for managing health care problems (or, less obviously, to select among health insurance plans). A second is to encourage specific changes in a patient's health-related
behavior. A third, as described in Chapter 5's discussion of risk management and medical liability, is to reduce the liability risks associated with poor communication and the disappointment that can result from unrealistic patient expectations.
Since the early 1900s, an evolving body of case law related to the crime of battery (touching without consent) has promoted ever-increasing attention to physician responsibilities for communicating with patients about the risks and benefits of proposed care and for obtaining informed consent to surgical procedures and similar interventions (Faden and Beauchamp, 1986; Mazur, 1988; Brennan, 1991a). The term informed consent itself dates from a 1957 California appellate court decision in Salgo v. Leland Stanford Junior University Board of Trustees (154 Cal. App. 2d 560 ); since then, courts have been trying to define what informed consent means and what it requires. In some states, the standard for judgment is whether a physician has disclosed what other physicians in good standing would disclose; in other jurisdictions, the standard is what a reasonable person in the patient's situation would want to know.6
Presumably, this judicial stimulus has increased the flow of information to patients, but current institutional procedures for informed consent seem intended, to a very considerable degree, to fulfill legal requirements and protect institutions from liability (Kapp, 1989; Hillman, 1991; Povar, 1991). This role is important. Nonetheless, a narrow, legalistic interpretation of the concept should not obscure the potential for informed consent to act as a vehicle for fulfilling patient preferences and improving the quality of care.
The President's Commission for the Study of Ethical Problems in Medicine and Biomedical and Behavioral Research (1983) noted that "although the informed consent doctrine has substantial foundations in law, it is essentially an ethical imperative [and] . . . a process of shared decision making based upon mutual respect and participation" (p. 20). The commission further argued that education and training, rather than judicial dictates, are the preferred vehicles for improving how physicians and others provide patients with the information they need. A focus on legal requirements can distract practitioners and institutions from the challenges that face profes-
sionals in making information useful to patients. Adding to the difficulties in this regard is that the ability of patients to comprehend and act on information may be compromised by emotional stress, psychiatric illness, intellectual limitations, financial constraints, and language and other barriers (Hillman, 1991; Povar, 1991).
These concerns noted, legal experts stress that informed consent does not exist by virtue of signed forms. If written (or oral) consent lacks real understanding, such empty agreement does not preclude liability. Put another way, legal requirements are consistent, rather than in conflict, with the ethic of informing patients, sharing decision making, and respecting autonomy.
As developers of guidelines become more cognizant of how variations in outcomes are perceived by patients and more specific about the risks and benefits of alternative courses of care for particular clinical situations, the guidelines they develop should provide a better base for patient information and decision making. Several challenges must be faced, however, in moving from initial guidelines documents, which are generally directed at clinicians, to patient-friendly guidelines and information. One challenge is the translation of risk-benefit analyses into messages that will register with patients both intellectually and emotionally. Considerable research has demonstrated the difficulties involved in creating realistic public appreciation of different kinds and levels of risk (National Research Council, 1989). Another challenge, consideration of patient preferences in the construction and use of guidelines, raises both technical and policy issues.
In the past few years, health services researchers have made clinicians and others increasingly aware that patients may vary in their preferences for different outcomes of care and that clinicians may perceive these preferences inaccurately (Wennberg et al., 1988; Mulley, 1989, 1991; Kaplan and Ware, 1989; Wennberg, 1990). That research has raised important questions about how to identify patient preferences, how to incorporate information about patient preferences into practice guidelines, and how to help patients make informed determinations about their preferences.
Reflecting a traditional emphasis on practitioners' obligations, discussions about health care decision making and patient preferences may not consider the ethical obligations and personal capacities of patients. On the one hand, the individual's personal and social responsibility for his or her own health behavior and choices may be ignored. On the other hand, how variability in individual intellectual, emotional, and other capacities may affect—and limit—patient decision making is a consideration sometimes lost in more theoretical discussions of patient preferences.
For developers of practice guidelines, incorporating patient preferences presents several challenges (Eddy, 1990d, 1991c). Guidelines developers will not have available much empirical evidence about patient preferences, lifestyles, and attitudes about different kinds of risks as these factors relate to specific clinical conditions or to health care generally. Securing such information will be expensive, even if the effort is aimed at the typical patient rather than the idiosyncratic patient whose preferences and behaviors may pose the greatest problem for clinicians. Further, how preferences or behaviors might or might not be accommodated in light of clinical evidence and judgment may raise significant ethical or policy questions that go beyond the expertise and responsibility of those participating in the development of particular guidelines (Granneman, 1991). This is not to say that guidelines should ignore the challenges of such behaviors and choices. Rather, it is to say that guidelines are unlikely to deal comprehensively with the totality of personal behavior and choice.
Yet even if guideline developers and users become adept at identifying and recognizing patient preferences, problems will remain for decision makers. Respect for patient autonomy does not dictate that physicians must always act to help patients or their families implement their preferences (Brett and McCullough, 1986; Povar, 1991). The controversy over continued care for patients in persistent vegetative states is a dramatic illustration. Patient demands for unindicated antibiotics are a routine but still troublesome problem.
Patient preferences may conflict with practice guidelines in at least three general ways, and these conflicts may raise minor to significant ethical questions. One kind of conflict exists when a patient demands care that appears not to be indicated according to a guideline involving the condition or service in question. Such a situation is least troublesome when the patient wanting the nonindicated service is willing and able to be responsible for any additional costs of providing it, is exposed to little or no risk, and imposes little or no burden on society. An example might be a low risk diagnostic test that is otherwise not indicated and that is paid for not by a third party but by the patient.
Other conflicts may involve patients who do not wish to receive care that is clinically indicated. For example, an athlete might prefer to play with an injury that normally would require rest, medications, or surgery. Another instance is the patient who refuses chemotherapy because the low expected benefit of such treatment (in her specific case) does not sufficiently outweigh (for her) its likely and unpleasant toxic side effects. Assuming that these patients have been fully informed of the possible consequences of exercising their preferences (such as a further disability injury or a shorter life expectancy), the two situations would also appear to offer little concern on ethical grounds. They do imply, however, that greater weight has been
given to patient autonomy than to some larger professional or social judgment of what is appropriate care.
More troubling are conflicts between a patient's preferences for care and a guideline when a third party (e.g., government, an employer, an insurer) is expected to absorb the additional costs if and when unindicated care is rendered. To the extent that third parties themselves use guidelines to inform utilization review and reimbursement decisions, they reduce the opportunities for patients to shift costs for unindicated care from themselves to others. However, to the extent that patients and physicians misrepresent clinical information in order to secure payment, the ethical problem is compounded.
Information, Preferences, and Policy: The Need for Guidelines on Patient Information and Informed Consent
The potential tensions among patient preferences, requirements for informed consent, and policies to contain health care costs are several. Even if policy makers somehow resolve questions about what care should be covered by private and public health benefit plans, other questions related to the provision of appropriate information remain.
Do all physicians have an equal responsibility to provide information about services that may have some benefit compared with alternative care, even when the more beneficial services are not available under the financing system or in the delivery setting in which they practice? If the answer is no, what information can be omitted? Under what circumstances?
More specifically, by enrolling in certain types of health plans, should patients forfeit their right to information—at the point of service—about treatment options of some benefit that are not (or may not be) covered by the plan? If so, is such a forfeiture absolute, or is it conditional on the provision to patients of clear advance warning that such limits may be applied?7 If the latter, who is responsible for that advance warning—government, an employer, the health plan, or the practitioner? How detailed should the warning be with respect to how limits are set and which specific
As noted in Chapter 5, health insurance contracts typically disclose certain kinds of restrictions. These restrictions may take the form of excluded services (e.g., cosmetic surgery), services limited by frequency (e.g., 20 mental health outpatient visits), or coverage that is conditional on patient compliance with preprocedure review and other utilization management requirements. The kind of disclosure considered here could involve a general statement that the plan reserves the right, for example, to apply certain practice guidelines or to employ a drug formulary. At one extreme, disclosure of specific protocols could be required. Alternatively, guidelines developed by specific organizations or according to specific criteria could be referenced (Havighurst, 1990b). Statutory and case law are still evolving in this area (Miller, 1991).
treatment options are not available (taking legal, organizational, and other practical issues into account)? Should the warning meet certain readability standards (e.g., eighth-grade reading level)?
What about government policies that forbid the provision of information, for example, on abortion or that require the provision of specific information, for example, on fetal development? What can and should practitioners do if they believe the regulations are scientifically or ethically improper? What proper role should government or other payers have in dictating what practitioners must and must not say to patients (Annas, 1991)?
The following discussion considers these questions only as they relate to the provision of information and the contribution that developers of guidelines might make in resolving these issues.
Provision of Information under Condition-Specific or Treatment-Specific Guidelines
This committee believes that developers of guidelines can do more than they do now to help practitioners define their responsibilities to provide information to patients. By describing the strength of the evidence for a particular guideline, estimating and assessing outcomes in terms that are relevant to patients, and more generally depicting how compelling is the case for different courses of care, they may guide judgments about how compelling are practitioner responsibilities for providing information and recognizing patient preferences. For purposes of illustration only, one possible hierarchy of obligations is outlined below. It is not endorsed in its entirety by the committee, in part because the assumptions about existing information are quite heroic.
First, when evidence and consensus are very strong, responsibilities to provide information likewise should be strong. This precept generally would hold even if the information concerned a service that was not available or not covered by insurance. A relatively obvious example is immunizations, which traditionally have not been covered by indemnity health plans but which clinicians routinely recommend, provide, or arrange for from public programs or other subsidized services. A more difficult case involves expensive services such as kidney transplants or dialysis, which some financing programs implicitly or explicitly ration on budgetary grounds.8
Second, when evidence is sufficient to support several treatment options with different mixes of risks and benefits, respect for patient preferences generally would prescribe the provision of adequate information to allow informed patient choice. Whether such a choice could be implemented (in particular, paid for by other parties through health insurance or subsidized service delivery programs) is a question that guidelines themselves can illuminate but not answer.
Third, when both evidence and consensus are unclear or nonexistent, the provision of information could be balanced against other factors such as time constraints and financial incentives. That is, if guidelines state that no evidence or strong consensus supports a particular course of care, practitioners face no compelling ethical (or legal) responsibility to provide information about that option.
General Guidelines for Patient Information
The committee also believes that a set of general guidelines for patient information and consent may need to be devised to supplement condition or treatment-specific guidelines, on the one hand, and legally oriented patient consent forms, on the other. Such guidelines would discourage an unsophisticated, narrowly legalistic approach to informed consent and confront the limitations of common mechanisms of disseminating information to patients (Green, 1991; Hillman, 1991; Povar, 1991; Siu and Mittman, 1991). They would be intended to provide specific assistance to clinicians, institutions, payers, patients and their surrogates, and any other involved parties in determining the types of information that should be provided to satisfy practical, ethical, and legal standards of care. Further, any broad set of information and consent guidelines would need to be relevant for (a) different kinds of care provided to (b) different kinds of patients in (c) different delivery systems and settings, given (d) different levels of certainty about the benefits, risks, and costs of care.
General guidelines on patient information and informed consent should be developed by a systematic process. Compared with processes for developing condition- or treatment-specific guidelines, the process for developing general information guidelines will involve somewhat different challenges and greater ambiguity. It is likely to call for greater consideration of ethical and perhaps other nonclinical factors in determining recommendations, demand more effort to define conceptual and operational measures of nonclinical benefits and risks, and require less specificity with respect to the vast number of individual patient situations that are likely to be encountered. These characteristics imply a more inclusive development process involving, among other things, more representation of health care purchasers and third-party payers, consumers and patients, and the legal profession.
Once formulated, general patient information and consent guidelines would apply to broad categories of patient care, unless they were specifically modified by condition-specific guidelines.
THE PHYSICIAN AS DECISION MAKER: WHAT CARE IS REQUIRED?
Good guidelines will be welcomed by physicians and other health care professionals. Nonetheless, such guidelines may still present practitioners with ethical problems. One committee member noted that he receives materials labeled ''guidelines" that call for a great range of services. His first question is whether any particular set of guidelines are, in fact, sound statements of what he really ought to do or recommend to preserve or improve his patients' health; his second question is whether practically he can do everything that is recommended; his third question is what hassles he may expect from payers, patients, or others if he does (or does not) follow the guidelines.
This physician's questions reflect the broader world of medical practice. Clinicians constantly make decisions and recommendations—some routine, some involving life and death—in the face of limited knowledge, time constraints, complex and unpredictable human behavior, and conflicting and even unreasonable messages from payers, courts, and others about the obligations of clinicians to patients and society (Morreim, 1989; Brook, 1991). The dilemma these conflicting pressures create is not always thoroughly appreciated by those outside the profession. That physicians are paid well does not negate the very real strains they may experience in juggling patient, payer, legal, and professional expectations, pressures, and disagreements.
To varying degrees, guidelines can help physicians and others by identifying how compelling is the case for particular services or courses of care under particular clinical circumstances. To further alleviate some of the strains on clinicians, some members of the committee argued strongly that developers of guidelines should specify the minimum (or basic or necessary) care required for each clinical problem or service they address. Such specifications would, in a sense, be intended to describe a "safe harbor," a statement of what physicians ethically and legally would be expected to provide to their patients. Negligence would be implied if they did not provide or at least recommend this minimum level of care.
Most committee members viewed this argument with sympathy, but the group ran into semantic, philosophical, and technical problems that prevented a clear consensus in the area. Nonetheless, the committee wanted to present some thoughts about how these issues might be debated.
One point quickly became clear in the committee's discussions. That is, although the issue of minimum, necessary, or basic care was first raised to the committee in the context of clinical practice and involved ethical,
legal, and practical concerns beyond third-party payment, the issue is most identified with controversies about what health insurance plans should cover. For some years, insurers, clinicians, and health services researchers have argued about what care is medically necessary or appropriate and who should make such judgments. That argument is taking on new intensity and significance as proposals for health care reform call for a package of basic benefits to be defined and used more or less uniformly by public and private health insurance plans.
Adjectives such as "necessary" and "basic" are quite common in everyday language, but they also have certain specialized uses that may be inconsistent with each other and with what might be termed ordinary usage. The result is a sizable opportunity for misunderstanding and failed communication. What follows is a brief review of definitions and perspectives intended to illustrate this point.
Dictionaries9 describe something that is necessary as being "of an inevitable nature," "compulsory," "absolutely needed," "required,'' "essential," "indispensable," "vital for the fulfillment of a need"; it is something "that cannot be done without" or that is "determined by force of nature or circumstance." Appropriate means what is "especially suitable or compatible, fitting," "suitable or fitting for a particular purpose; proper," or "specifically fitted or suitable." What is indicated may be "necessary" or, less strongly, "advisable" or "suitable." That which is basic is "fundamental," "essential," "constituting the starting point," "primary," or "of lowest rank." A minimum is the "least quantity assignable, admissible, or possible" or the "least amount attainable, allowable, or usual."
The term medical necessity appears to have arisen several decades ago as newly developing health plans sought to limit payment or reimbursement to only that care that was medically necessary for the diagnosis and treatment of a condition, illness, or injury.10 However, not all insurance programs employ the term in contracts and elsewhere, and not all programs that use the term actually define it. When definitions are provided, they vary considerably (Helvestine, 1989).
The definitions here are drawn from Webster's Ninth New Collegiate Dictionary. the Random House Dictionary of the English Language, and the Compact Edition of the Oxford English Dictionary.
This terminology inspired the name and the purpose of the original Blue Cross Medical Necessity Program, which was described in Chapter 2. The fact that the American College of Physicians, which was asked by Blue Cross to assist in assessing medical necessity, named its program the Clinical Efficacy Assessment Program suggests that the medical profession was not completely comfortable with the former term, at least in this context.
The least restrictive definition seems quite simple: medically necessary care is the care a physician provides or prescribes. This definition excludes payment for nonmedical treatments such as faith healing but otherwise defers to the physician. Medicare regulations go a little further, referring to care that is safe and generally accepted by practitioners. Those health plans that use Value Health Sciences systems may be adopting, implicitly if not explicitly, the RAND definition of appropriate care—that is, that the medical benefits of a service exceed medical harms by a sufficient amount to make the service worth providing. In any case, despite the shift by insurers away from unfettered physician discretion in determining medical necessity, the actual interpretation of medical necessity seems to fall between the dictionary meanings of necessity and appropriateness rather than to follow the definitions of the former. A fuller examination would likely show that interpretations vary considerably; for example, newer technologies may well be treated more strictly than older technologies, and matters of site and timing of care may be subjected to more questions than the matter of whether to provide that service at all (IOM, 1989a).
The RAND definition of appropriateness is useful to cite because it more readily prompts the question that underly all the definitions: How much benefit is enough for care to be rated as appropriate (or necessary)? Is it any possible marginal benefit? Is it what physicians say is an important benefit? Is it what an individual patient or the average patient thinks is enough benefit? Any such judgments and distinctions are highly subjective and reflect, in part, the different concerns or interests of different parties.
At least six special concerns seem to be discernible in discussions that employ the terms defined above.
The public or private insurer perspective. It asks, What health services should we pay for? What services will courts say we must cover?
The professional liability or risk management perspective. It asks, What care will the courts hold practitioners responsible for providing or recommending?
The perspective of the ordinary person. It asks, Am I going to get the care I need and want? Will it be paid for?
The evidence- and outcomes-based perspective. It asks, What does science say about what works, and how convincingly does it say it?
The conservative style of practice perspective. It asks, When is intervention required rather than watchful waiting?
The ethical perspective. It asks, What does justice or decency require society to assure its members?
These concerns are by no means mutually exclusive, but neither are they identical. Moreover, within the context of a single perspective, differ-
ent attitudes can be distinguished about what constitutes a sufficient case for medical intervention.
Adding further to the confusion about adjectives such as necessary, basic, or minimum is the fact that the noun care is sometimes not clearly distinguished from such terms as coverage or benefits. The latter refer not to health services per se but rather to payment for services under the terms of a health benefits plan.
In conventional insurance terms, one is insured against a financial loss caused by some peril; in the case of health care, the loss is the money spent for medical services incurred as a result of illness or injury. The benefit is what the insurer contributes to meeting those losses. Losses and benefits are ordinarily specified in considerable detail in insurance contracts, which may restrict benefits to certain settings of care, types of practitioners, medical conditions, and so forth. "Health care benefits" clearly are not the same as the "benefits of health care" in that some covered care may not be beneficial and some beneficial (necessary or appropriate) care may not be covered. Cases in point involve the not uncommon exclusions of coverage for immunizations, blood products, and dental care; Medicare also excludes outpatient prescription drugs.
In discussions of broad principles for health care delivery and financing, the term basic benefits clearly has a variety of different meanings (Veatch, 1991; Hadorn and Brook, 1991; see generally the May 15, 1991, issue of the Journal of the American Medical Association). Some meanings are implicit, some explicit. Some are consistent with the common meanings associated with the term basic, that is, something that is "essential" or "of lowest rank," "primary'' or "constituting the starting point." Other meanings are not in line with common usage and may even be misleading. Among the varied conceptions of basic benefits are the following.
By their listing of basic benefits, some health care reform proposals seem to mean simply the ordinary kinds of health care services, settings, and providers covered in the typical (middle-class) health plan. Basic in this sense dates back to early health insurance (primarily Blue Cross) terminology that described fully covered hospital and physician services as basic and other services as supplemental. The term standard benefits might be a better label for this conceptualization of a benefits package.
Other discussions suggest that a basic benefits package is an "urgent care" or perhaps "bare bones" package aimed primarily at the kinds of illness or injury that produce significant expenditures (e.g., above a relatively high deductible) but not necessarily catastrophic expenditures (e.g., more than 30 or 60 days of hospital care).
In contrast, some appear to see basic benefits as those involving preventive and primary care services that have relatively low unit prices and simple technology (e.g., immunizations, well-baby care).
A few conceptualizations start with the relatively broad range of services now covered by most health plans but then attempt to limit coverage to "effective" services based on considerations of evidence, cost, importance to individuals, and social value.
This last perspective is reflected in some broad proposals for reforming this nation's health care financing and delivery system (see Chapter 8). This perspective also is reflected in Oregon's Basic Health Services Act, which directs a revision of the state's Medicaid program (Eddy, 1991d; Granneman, 1991; Hadorn, 1991c). The legislation calls for services to be ranked by priority and then covered in order of ranking until the program budget is exhausted. The commission charged with setting priorities has employed town meetings, quantitative analyses, subjective judgments, and various other processes to generate information about costs, outcomes, and individual preferences for particular outcomes. It has then used this information in setting and revising priorities. Arguing for an alternative but still comprehensive approach, Hadorn and Brook (1991) would rely solely on judgments of health benefit excluding judgments of cost-effectiveness. Under their approach, a basic benefits package would cover only services that provided significant net health benefits; these services would be not just effective, beneficial, or appropriate but "necessary to a minimally decent life."11
Clearly, terminology in this area is confused and fraught with ethically and politically sensitive connotations that intensify the impact of any misunderstandings. For convenience in considering practical and policy issues, the next section of this report employs the term minimum care, a term the committee thought carried less policy or political history than the others discussed above. Minimum care in this discussion is not a matter of requiring high deductibles and cost-sharing in health insurance plans but involves what specific services are covered by such plans.
Practical and Policy Issues
Even if a term such as minimum care is agreed upon, many difficult operational and policy questions confront efforts to specify just what constitutes such care. One practical issue is whether to attach the term minimum only to those individual elements of care that are strongly based in
scientific evidence and that developers of guidelines can, indeed, describe as required under most circumstances.
This evidence-based approach to defining minimum care has serious limitations. For many conditions for which no strong evidence exists, no specific care could be deemed to be required, even though the alternative—no intervention at all—might be inconsistent with available evidence or strong expert consensus. The blood transfusion example cited in Chapter 1 is a case in point; no direct evidence exists about the precise threshold at which transfusion is indicated, but no one would counsel that transfusions are not required at some point to avoid death or injury in a variety of situations. Similarly, most screening guidelines acknowledge that even when research supports a particular screening service, the evidence is unlikely to speak to the particular interval for screening. An equally serious problem was noted in Chapter 1: scientific evidence is not likely to exist for a great many of the combinations of clinical problems and characteristics that patients bring to clinicians in the real world.
Minimum care would probably need to be defined as a constellation of services, for example, the least number of services (or options) for managing a condition that could be supported by strong expert consensus and that was consistent with available scientific evidence. Unfortunately, trying to draw lines around sets of services will be an even more subjective and controversial process than trying to draw the line for or against a specific intervention. Efforts to assess patient preferences and reflect them in global coverage policies (as attempted in Oregon) have been criticized as inherently unable to deal adequately with variations in individual needs and values (Granneman, 1991).
Efforts to define minimum care and set priorities for insurance coverage across the entire array of existing health services may run into additional challenges not faced by more incremental strategies. Collecting and analyzing information and making objective and subjective comparisons involving thousands of services and combinations of clinical circumstances constitute such a monumental undertaking that simplifying strategies inevitably arise. These strategies—for example, grouping services together in broad categories—may be methodologically flawed and may compromise the resulting judgments (Eddy, 1991d). Although the defects may be fixable in principle and the fixing may be doable for a limited number of services, it is not clear that they are feasibly applied comprehensively to all or even most services.
An alternative, incremental approach would concentrate on "ruling out" ineffective services rather than "ruling in" only effective ones. It would focus on such "targets of opportunity" as new and emerging technologies, obsolete services, services characterized by wide practice variations or thought
to be overused or misused, and services that otherwise are of keen interest to policy makers, practitioners, and patients. This is essentially the approach now employed by Medicare and other payers. Some members of the committee believe that incremental approaches are the only workable ones (albeit in need of more serious commitment of resources available now); comprehensive strategies promise far more than they deliver— technically, administratively, and ethically. In this view, the Oregon initiative is a valuable exercise but not a policy model.
In addition, some worry that efforts to define minimum care would become an operation to describe not a floor beneath which care should not fall but a ceiling beyond which it should not rise. They also fear that such distinctions will preclude "excellent" care or will compromise a physician's sense of responsibility for a particular patient whose circumstances might justify more or different care. Whether other-than-minimum care should be defined as excellent care, however, raises questions about whether excellent care is to be distinguished by better expected health outcomes, better accommodation of patient preferences, or something else. More care, in and of itself, is not necessarily better care, although it certainly may be.
Other important questions face any effort to identify a constellation of minimum services. For instance, whose "minimum" is at stake? Whose perspective—that of an individual or a population, or of a practitioner, patient, or policy maker—should govern in establishing that minimum? Should the same minimum apply for all purposes? For example, should the same minimum determine what care is to be insured (or made available to all) and, at the same time, serve as the standard for determining negligence? Should people be required to receive minimum care in some circumstances or risk losing insurance or other benefits, as has been suggested recently for welfare recipients or others in Delaware and Maryland (Goldstein, 1991; Robb, 1991)? Can any single process for defining minimum care accommodate the differences in incentives among systems of care, for example, fee-for-service and capitated systems (Granneman, 1991)? Should the same minimum apply to those covered by Medicaid and those covered by employment-based programs? Considering the different circumstances of the "average" poor person in the United States and the average individual in an impoverished developing country, will the definition of minimum care be bound as much by culture and resources as by evidence?
Opinions clearly differ on these issues and reflect complex differences in value judgments. Some argue that those who advocate explicit identification of minimum or basic care must be prepared to accept that minimum for themselves in, say, a basic insurance benefit package provided or subsidized by government or employers. Others disagree. Furthermore, some
who agree about the basic package for subsidized insurance disagree about whether it would be unethical for them to supplement the package by paying for additional care that others might want but are unable to afford. Finally, whether one has to resolve these ethical questions before arguing for statements of minimum care is itself a matter of dispute.
A final point: what started out as a committee discussion of minimum care from a clinician's perspective became a discussion of minimum care from an analyst's perspective. The results of an effort to define minimum care for public and private health plans might make it easier for physicians to predict what care would be paid for; however, it is likely to leave unresolved (or to complicate) a variety of other ethical, practical, and legal issues that concern clinicians.
What Should Developers of Guidelines Do?
Given the terminological, practical, and ethical problems raised by the issue of minimum care, the committee confronted this fundamental question: Can and should the (relatively) fragile enterprise of guidelines development be expected to take on extremely sensitive and highly complex issues of "valuation" of health care services for society at large? After extended debate, the committee concluded—with some dissent—that the answer is no. It is not prudent to recommend that guideline developers uniformly state the minimum, necessary, or basic level of quality care for every clinical problem or service for which guidelines are formulated. However, guideline developers should attempt to describe the incremental benefit associated with particular courses of care.
Some developers of guidelines may be technically, ethically, and politically positioned to propose minimum care for a limited set of clinical problems, but many others do not now and may never wish to assume this responsibility. More fundamentally, developers of guidelines do not appear to this committee to be the appropriate locus for declaring what is minimum or basic care insofar as those decisions apply to third-party payment or broader resource allocation policies. Demanding that guidelines developers be explicit about "minimums" may undermine the credibility of the entire process, make the evaluation of the science base extremely vulnerable to political biases, and reduce the process, in the view of some, to an exercise in defining "two-tier" health care.
The committee does wish to express its discomfort with the terminological confusion, even sloppiness, surrounding the use of such phrases as "medically necessary care." This discomfort does not stem from a desire for linguistic perfection. Rather, it arises from concerns that very important decisions are being made on the basis of poorly defined criteria, a process
that will result in inconsistent and often conflicting judgments that, in turn, will induce confusion, hostility, and, ultimately, inequity.
Management (if not resolution) of the tensions discussed in this chapter concerning ethics, costs, and information will depend on decisions about how health care is to be financed and delivered in the future. Developers of guidelines can illuminate debates over various individual and collective interests by presenting evidence, analysis, and expert judgment about the risks, benefits, costs, and patient preferences associated with alternative courses of care. Well-developed, evidence-based guidelines that are specific, logical, clearly explained, and accompanied by projections of health and cost outcomes (to the extent possible, given the dearth of this kind of information) can and will be incorporated in quality, cost, and liability management programs. Their incorporation, in turn, will provide powerful support for the consistent application of such guidelines in actual clinical practice.
Nonetheless, differences in philosophies, resources, attitudes toward risks, and other factors will ensure some inconsistency and dispute. Clinical experts and decision makers may argue among themselves about how to interpret weak or conflicting scientific data, how to estimate and weigh benefits, harms, and costs, and how to resolve questions of individual versus collective perspectives. Further, pure objectivity and perfect rationality may exist in the realm of theory but not in the world of real human endeavors. Decisions to use or not to use particular guidelines may consciously or unconsciously reflect economic considerations, inclinations toward "conservative" or "aggressive" styles of practice, and other factors. This is one reason that the committee places such emphasis on the attributes of good guidelines—they should reduce the opportunity for important but unacknowledged values or biases to affect the formulation or application of guidelines.
The committee judged that it is not now strategically or tactically prudent to impose on all developers of clinical practice guidelines the task of explicitly recommending what care is warranted on economic as well as clinical grounds. Nor should guideline developers be uniformly expected to declare what services constitute the minimum or required care for a clinical problem. As important and necessary as these judgments may be, developers of guidelines for clinical practice need not take on this responsibility.
A fundamental reason for this position is that users rather than developers of guidelines carry the actual responsibility for deciding how to
deploy resources and how the projections of health and cost outcomes offered by guidelines relate to their specific circumstances and objectives. A second, practical reason for the committee's position is that developers of guidelines may be nearly overwhelmed in responding to the expectations already laid out for them in this report and elsewhere. In the committee's view, it is important that they concentrate on these tasks and show that they can build a credible foundation for better decisions by practitioners, patients, and others. The next chapter pursues this point.