Click here to access the corrected complete pdf version.
Opening the black box of clinical judgment
Part IV: An Economic Perspective
Lincoln Weed, Lawrence L Weed
Readers sympathetic to the reforms we have outlined may nevertheless be concerned that the economic payoff is too uncertain and the change involved too great for the reforms to be feasible. In part V we explain why reform is more feasible than it may first appear. This part examines the economics of the present health care system and some of the economic effects that reform might have. We begin with some analysis and evidence of the adverse economic effects of uncontrolled inputs. Then we discuss the economics of quality in light of the theory of competitive markets, arguing that the new tools and approaches needed to control inputs can reduce dependence of health care consumers on providers for medical knowledge and judgments of quality. We then argue that equipping consumers to function autonomously would release market forces in a way that advances both medical and economic goals in privately and publicly funded health care.
A. The high costs of low quality in medical practice
The extent of avoidable error and waste in medicine has yet to be fully grasped, for many reasons. Dramatic scientific advances dominate awareness. Dedicated and talented health care workers apply medical knowledge as best they can and succeed more often than not. Flawed tools and premises block understanding of the nature and extent of failure. Serious medical errors thus seem to be occasional, isolated events, the inevitable residue of human fallibility in the face of enormous complexity.
Yet, tolerating even isolated instances of error can increase costs exponentially. This occurs at multiple levels:
• When an inexpensive medical procedure is also ineffective, its low cost represents economic loss, not savings.
• Ineffective medical procedures are not only wasteful in themselves but typically escalate costs, because an expensive process of trial and error is set in motion.
• Erroneous or unnecessary procedures can further escalate costs by harming the patient. Medical procedures may be executed improperly, or may have potential for dangerous side effects, interactions and other complications. Further risky medical intervention may thus be triggered.
• Costs can also escalate when a missed diagnosis or incorrect treatment delays effective treatment, allowing the patient’s condition to worsen.
• The economic ripple effects of poor quality medical decision making are incalculable. Resources are diverted to ineffective medical care from more productive uses, and patients are delayed from returning to productive activity. Poor use of existing medical knowledge generates distrust of providers, unnecessary payment disputes and malpractice litigation, which dissipate resources and lead to more decisions of poor quality (e.g. "defensive medicine" by providers, misplaced reliance on alternative medicine by patients). Missed diagnoses and ineffective or harmful treatment decisions can be emotionally harmful to patients, who may be wrongfully branded with psychological diagnoses, as the Addison's disease case illustrates. Such outcomes disrupt family and work relationships and further reduce the economic value of medical care.
In short, the cost of error may far exceed any immediate savings from tolerating error. And tolerance of error on a large scale is built into everyday medical decision making, with its dependence on the unaided human mind. Moreover, even when outright error does not occur, care is often not justifiable in terms of risk, need, cost, benefit or patient wishes. Again, this kind of quality failure is built into everyday medical decision making, with its failure to inform and involve the patient. It is thus unsurprising to find that quality failures are not isolated but pervasive. Their persistence in different economic environments shows the need for fundamentally new tools and approaches for controlling inputs in routine patient care.
We have focused on cognitive inputs, which are crucial to exploiting information resources in medicine. Economists recognize that information, because it can be freely copied and distributed, need not be subject to scarcity constraints. The primary constraints on exploiting information resources are timeliness and usability. These constraints mean that the tools and processes by which users retrieve, organize and communicate medical information are pivotal. Tools to exploit information resources to the fullest must be put to use in every medical encounter with every patient. Until this happens, the achievements of medical science for many patients will continue to go to waste. As described in a recent discussion of the relation between costs, quality, and uncertainty in patient care:
Clinical medicine continually confronts clinicians with their own ignorance, which may be difficult to distinguish from deficiencies in medical knowledge as a whole. . . . The expansion of medical knowledge, rather than diminishing uncertainty, is contributing to its increase as doctors’ ability to process the mushrooming volume and complexity of data is being outstripped at an accelerating rate. (1)
Existing information tools such as telemedicine, on-line research sources and written practice guidelines are inadequate to change this state of affairs. Telemedicine simply extends the geographic reach of conventional decision making. On-line research sources, while they to some degree address the problem of timeliness, do little to solve the problem of usability, and can even exacerbate it. By accelerating retrieval of general medical knowledge, on-line sources increase the mind's burden of organizing and evaluating knowledge from multiple sources, and integrating that knowledge with detailed patient data. Asking a busy practitioner to perform these functions is like giving a hungry person a bushel of wheat rather than a loaf of bread. Practice guidelines in their current form are little better. In theory, practice guidelines serve the function of consolidating and evaluating medical knowledge. Even when they achieve this goal (and they often do not), guidelines still leave it to the physician to integrate general knowledge with patient-specific data. That processing function is beyond the mind’s capacities under the time constraints of actual clinical practice. Other problems limit the practical utility of paper-based guidelines as well. Not surprisingly, therefore, dissemination of practice guidelines has been found insufficient to change physician practices. (2) (3) (4) In the words of one practitioner, Dr. Charles Burger, leading guidelines developed at great expense by the Agency for Health Care Policy and Research "now lie dormant in desk drawers all over the country. New guidelines promulgated by various managed care companies arrive regularly in bound notebooks to suffer a similar fate." (5)
In short, the health care system is designed around an archaic belief that the physician’s trained intellect may be relied upon to manage medical information. The harmful consequences of this belief are especially apparent in two areas with great economic impact — the U.S. Medicare program, and drug therapy.
The Medicare population has a disproportionate share of patients whose cases are difficult to comprehend, because the elderly tend to have multiple, interacting problems and treatments. Therefore, the phenomena illustrated by the Addison’s disease case study — information overload, physicians’ cognitive limitations, risk to patients, economic waste — are especially likely to occur with Medicare patients. Seen in this light, it is not surprising that, as the Congressional Budget Office has observed, two thirds of recent growth in Medicare spending is attributable not to increases in population or provider payment rates but rather to "unexplained increases in the volume and intensity of services." (6) One likely explanation for such increases is that doctors lack the information tools needed for cost-effective medical decision making when faced with an ever-expanding knowledge base and a growing arsenal of complex, risky medical interventions. Another explanation, more generally recognized, is that fee-for-service doctors increase medical services to compensate for lower Medicare reimbursement rates. Decision making on that basis itself represents poor quality. Thus, both explanations suggest that new tools and approaches for medical decision making are essential if Medicare costs and quality are ever to be brought under control. And the untapped potential for cost control in Medicare may well be large. In addition to the high costs of avoidable error discussed above, improved quality might help further reduce the elderly's disability rates, which could have a substantial moderating effect on costs. (7)
Drug therapy further illustrates the need for reform. Analyses of the literature by the U.S. Food and Drug Administration and the U.S. Congress' General Accounting Office indicate that large numbers of Medicare beneficiaries (estimates ranged from 14 to 25 percent) took at least one of a list of 20 drugs generally considered unsuitable for the elderly and for which better alternatives exist. The General Accounting Office noted that these figures understate the scope of the problem, because other incorrect drugs and other types of inappropriate drug use (such as dangerous interactions and incorrect dosages) were not considered. The General Accounting Office also cited studies estimating "that 32,000 senior citizens annually suffer hip fractures as a result of falls caused by adverse drug reactions" and that "about 16,000 car accidents resulting in injuries each year can be attributed to adverse drug reactions experienced by elderly drivers."8] Other studies not limited to the Medicare population are consistent. The FDA estimates that hospital admissions for preventable adverse drug reactions alone cost $4.4 billion annually. (9) More recent studies, extrapolating from experience at individual institutions, reach conclusions of the same order of magnitude, and these cite studies estimating that in the United States the total annual cost of drug-related outpatient morbidity and mortality is in the neighborhood of $70 to $130 billion annually. (10) (11) (12) Indeed, error rates may be increasing. (13)
The studies often do not differentiate between decision making and execution of decisions relating to drug therapy. It is clear, however, that a large proportion of preventable errors in drug therapy occur due to failures of information processing at the decision making stage. (14) (15) (16) Moreover, a recent economic study provides evidence that physician attributes such as exposure to advertising influence prescribing decisions. (17) Some clinical pharmacology professors believe that the answer is to train more clinical pharmacologists, which requires two to three years of sub-specialty training after seven years of medical school and specialty training. (18) Such an approach is not only prohibitive in time and expense; it is futile. As explained by one of the authors of the 1995 studies cited above, "[t]he amount of information that surrounds the drug-prescribing process is enormous. For any one individual to keep track of it all . . . isn't realistic. The information system needs to be improved." (19) This is true even within a narrow specialty such as psychopharmacology. (20) Improved information systems for drug therapy have indeed been developed. (21) These realities further illustrate the absolute necessity of escaping dependence on the human mind to retrieve and process medical information.
Similar conclusions apply even for unforeseeable adverse reactions to drugs that are thought to be correctly prescribed and administered. Serious adverse reactions occur so frequently that a debate now exists about whether they should be considered among the leading causes of death in the United States. (22) (23) (24) Moreover, once a patient experiences an adverse reaction, "avoiding administration of the same medication to the patient in the future requires knowing and documenting that the patient had a previous allergy or sensitivity. When a patient develops an allergy or sensitivity, it is often not recorded, and patients receive drugs to which they have known allergies or sensitivities with disturbing frequency." Furthermore, rigorous recording and review of clinical experience with medications is necessary to learn more about their risks for all patients. This cannot happen with current medical records, which are too haphazard for review to be effective or affordable, and current reporting systems, which reportedly only identify about five percent of adverse reactions. (23)
B. A theoretical perspective.
Civilization is an enormous device for economizing on knowledge. . . . The march of science and technology does not imply growing intellectual complexity in the lives of most people. It often means the opposite. . . . Organizational progress parallels that in science and technology, permitting ultimate simplicity through intermediate complexity. — Thomas Sowell (25)
We now turn to the economics of quality in light of the theory of competitive markets. Kenneth Arrow analyzed the limitations of market theory for health care in a seminal 1963 article. (26) The theory of the ideal competitive market is that — given an initial distribution of purchasing power and given existing resources and technology — voluntary, self-interested transactions among buyers and sellers allocate resources optimally (no other allocation would result in greater aggregate welfare among marketplace participants). The health care market, however, does not achieve this state of optimality "due to the nonmarketability of the bearing of suitable risks and the imperfect marketability of information." The result is market failure, that is, "a reduction in welfare below that obtainable from existing resources and technology." Market failure gives rise to compensatory, nonmarket social institutions. Licensing restrictions on medical practice, for example, give assurances of quality (in theory) to patients who depend on practitioners for information, judgment and skillful performance.
Our analysis of medicine's design flaws suggests that market failure is even greater, and at the same time more avoidable, than is generally understood. Focusing on the realities of poor quality in medicine makes it possible to view market failure — "reduction in welfare below that obtainable from existing resources and technology" — in a new light. Rational choice theory leads economists to assume that, given the proper incentives and information, producers use their rational capacities to achieve the best results "obtainable from existing resources and technology" while consumers use their rational capacities to judge quality and make cost-quality trade-offs. Economists then contrast health care markets, where consumers have limited information, with other markets, where consumers have sufficient information to exercise rational choice. This contrast, however, misses the point that in health care not only consumers but producers themselves lack sufficient cognitive capacity to exercise rational choice effectively. As discussed above, cognitive limitations disable doctors from achieving the best results obtainable from existing resources and technology. Until effective information tools are employed to bypass cognitive limitations, the health care system has no chance of approaching the theoretical ideal of competitive markets.
This perspective does not appear to be generally recognized in economics and other disciplines that rely on rational choice theory. Even critics of rational choice theory such as E. O. Wilson take cognitive constraints as a given: "The human brain is not a very swift calculator, and most decisions have to be made rather quickly, in complex settings and with incomplete information. So the question of importance in rational choice theory is, how much information is enough." (27) How much information is enough, however, is a function of what information the problem at hand requires and the effectiveness of available means for retrieving and processing information. To the extent that tools external to the mind accurately identify relevant information and filter out the irrelevant, then often we can set a higher threshold for how much information is enough and a higher standard for reliability in decision making.
The availability of information tools calls into question the usual assumption in health economics that consumers of health care depend on providers for medical knowledge, judgment and skill. That assumption has long been questioned from a variety of perspectives (28) (29), and now the availability of new information tools has led to further questioning.
Kenneth Arrow's influential statement of why consumers are dependent on producers in the health care market centers on the phenomenon of uncertainty. The prevalence of uncertainty, he argues, makes knowledge or information into a commodity in demand by consumers.
But the demand for information is difficult to discuss in the rational terms usually employed. The value of information is frequently not known in any meaningful sense to the buyer; if, indeed, he knew enough to measure the value of information, he would know the information itself. But information, in the form of skilled care, is precisely what is being bought from most physicians, and, indeed, from most professionals. (p. 946)
Moreover, unlike other commodities, where the buyer's knowledge of the utility of his purchase is as good as the seller's knowledge, in medicine the seller has superior knowledge concerning the utility of medical services, according to Arrow (pp. 951-52). Thus, patients are dependent on physicians for information to an extent that does not occur in most buyer-seller relationships.
We depart from Arrow's analysis in two respects. First, we distinguish between knowledge and skill in the seller. At the time of Arrow's article (1963), equating the two as he did was not surprising, because of the apparent absence of alternatives to physicians as a source of information. Now, however, the very information tools and medical records that doctors need to manage information can also give patients usable access to information specifically relevant to their medical needs. The physician's superior general knowledge is not relevant from an economic perspective, because, like other consumers, patients only need the information specifically relevant to their particular decisions. And the doctor need not be the source of that information.
Second, we depart from Arrow's assumption that the physician's knowledge of the utility of medical services is superior to the patient's knowledge. Because of the importance of the patient's self-knowledge of his own condition, and the inherently personal nature of uncertain medical decisions, the patient's knowledge of the utility of medical services may be superior to the physician's, once the patient becomes informed with personally relevant information needed to judge utility.
For these reasons, what patients need to buy from physicians is not information and decision making but high skill levels in executing medical decisions. This conclusion may seem to ignore marketplace reality, but that reality simply reflects a flawed health care system. Economist Thomas Sowell has explained a different reality that the present health care system ignores:
Decision-making processes differ not only in the quantity, quality and cost of knowledge brought to bear initially, but also and perhaps still more so, in the feedback of knowledge and its effectiveness in modifying the initial decision. This feedback is not only additional knowledge but knowledge of a different kind. It is direct knowledge of particulars of time and place, as distinguished from the secondhand generalities known as "expertise." The high personal cost of acquiring expertise, and the opportunities it presents for displaying individual talent or genius, make it a more dramatic form of knowledge, but not necessarily a more important form of knowledge from a decision-making point of view. (30)
In many economic contexts other than health care, the limited relevance of expert knowledge to personal consumption decisions is taken for granted. One need not hire an engineer to buy a car; one need not hire a geography professor to learn the route for driving from New York to Atlanta. Market forces have evolved systems enabling consumers to function in such activities without hiring expert professionals. Similarly, in medicine, tools for reliably processing complex information can simplify the ultimate choices presented to consumers by filtering out irrelevant information while clearly presenting relevant options and the pros and cons of each based on the consumers' individual characteristics.
C. Reform and the operation of market forces in health care
The reforms we have described would make it possible to for market forces to operate much more effectively in health care than is usually thought possible. Consider first credentialing — in Arrow's terms, a nonmarket social institution thought to give consumers assurances of quality. Information tools will usually not be sufficient for patients to feel comfortable in making major medical choices on their own. Patients will often need and want a dialogue with another human being, both to help understand the information presented by software and medical records, and to learn subtleties that those information tools cannot capture. What qualifies caregivers to engage in dialogue with patients about their medical decisions is not what credentialing requires — formal education and training — but rather interpersonal skills, use of the necessary information tools and concrete experience in caring for other patients with similar medical problems. Interpersonal skills do not require credentialing; market forces can be relied upon to select among caregivers in that regard. Nor is credentialing required in the domain of knowledge, where information tools rather than erudition should be relied upon to assure quality. Credentialing is needed simply to give consumers assurance of high skill levels in carrying out medical procedures.
Traditional credentialing does not assure skill, indeed, it interferes with attainment of high skill levels by physicians who expend their limited capacities in Sisyphean efforts to master medical knowledge. Moreover, traditional credentialing harms patients who lack access, for economic or geographic reasons, to expensively educated professionals. In contrast, rigorous credentialing of caregivers based on demonstrated skill would have several marketplace benefits. First, it would reduce the enormous time and cost of medical education. Second, it would remove a barrier to entry for those who could acquire and demonstrate competence in specific caregiving functions without passing through an expensive educational process. A free, competitive market in health professional services could thus develop. Third, this competition would not undermine quality. The rigor of credentialing based on demonstrated competence would give consumers and third party payers assurance that choosing a lower cost provider does not entail an unacceptable trade-off against quality.
Giving consumers this assurance, combined with giving them access to information tools, would enable them to function more autonomously than ever before. That development could go far to remedy an overlooked form of market failure in health care. The present marketplace permits the law of diminishing returns to operate with little restraint, so that medical care often yields no benefit and may cause harm, as Elliot Fisher and Gilbert Welch have recently shown in an illuminating analysis. (31) Some of the mechanisms they describe — identifying subtle abnormalities with no clinical effects, lower diagnostic thresholds for defining disease, risky treatment for inconsequential disease, futile treatment, treatment in response to normal physiologic variation — would be much less likely to operate if patients were better informed and participated more fully to provide effective feedback in decision making.
The concept of market failure is usually applied to the financial aspects of the health care system, and there too consumer autonomy is critical. Fee-for-service providers would have little opportunity to create demand for unnecessary services when patients (and third party payers) can readily use software tools and medical records to examine the clinical justifications for services. And capitated providers tempted to withhold needed care could not easily do so, because information tools would objectively identify the available options and their pros and cons. Consumers can thus be protected from the financial conflicts of interest that arise whenever provider compensation is linked to provider decisions.
Indeed, with respect to capitation, it seems doubtful that such a form of compensating providers would continue in the long run if providers were equipped to provide care cost-effectively and if patients were equipped to participate in decision making. Capitation evolved to combat fee-for-service abuses by giving providers financial incentives to economize. The advantage of capitation for providers is to avoid third party micromanagement approaches to economizing, and to reward providers who employ new tools and approaches effectively to increase their productivity {and minimize expensive referrals). The disadvantage of capitation, however, is to create direct conflict between the financial interests of providers and patients when expensive care is unavoidable. This conflict gives rise to perverse incentives — doctors are tempted to avoid the sick and withhold care from their patients who become sick — that would not arise in a properly functioning market, where sellers compete for buyers by improving cost and quality. (32) The raison d'etre for such perverse incentives largely disappears once providers and providers are equipped to function like other economic actors. Indeed, capitation would be counterproductive where efficient caregivers respect patient autonomy in decision making, because then cost variations reflect variations in patients' needs, not in provider conduct.
Some readers, however, may believe that capitation is needed on the ground that patients are insured and lack incentives to economize. Such a view mistakenly assumes that fully insured patients have as much incentive to obtain unnecessary care as fee-for-service providers have to supply it. The reality is that for patients medical care is an unfortunate necessity, not a source of income. Fully insured patients thus have an incentive to obtain the best care but no more. And the best care is sometimes not the most expensive care, especially from the perspective of patients, who weigh the personal burdens that expensive care often entails. It is true that not all patients are so restrained; some "chronic patients" seek unjustifiable medical care to satisfy emotional needs. But the solution to this problem is not to concentrate decision making power in the provider at the expense of patient autonomy (which is the effect of capitating provider compensation). Rather, the solution is giving autonomous parties ready access to objective information, and, when necessary, a practical system of adjudicating disputes. Patients, providers and third party payers should not be able to impose medical or financial decisions on each other unilaterally. Checks and balances are needed.
Equipping patients to participate in medical decisions creates checks and balances that diminish the need for third party micromanagement of medical decision making. Without such reform, the health care system will continue to be crippled by what Robert Kuttner has called "an endless regress of private regulation — multiple and often overlapping sets of practice guidelines, quality audits, utilization reviews, physician profiling, and at-risk incentive formulas modified by appeals systems and ethical constraints — as well as multilayered government regulation and litigation." (33) Further examples of such complexity are "risk adjustment" and quality measurement, which are mistakenly thought to improve the operation of market forces.
Risk adjustment at the provider level is intended to mitigate the failings of capitation. On this view, providers saddled with more expensive patients receive correspondingly greater capitation payments but still have incentives to economize. Yet, it is not clear that such risk adjustment can ever be achieved with sufficient precision, especially given the inadequate state of current medical records, and the inherent imprecision of the textbook disease classifications used to adjust risk. Moreover, even the most precise risk adjustment is limited in its ability to predict the distribution of health expenditures, making it inevitable that providers will be overcompensated or undercompensated to some degree. Finally, capitation inherently gives providers a short-term financial incentive to avoid expensive care; risk adjustment can do nothing more than change the baseline. Problematic incentives and enormous complexity are thus inevitable under capitation. (34) It is true that fee-for-service compensation does not escape this dilemma, because of its inherent bias towards unnecessary services, but at least fee-for-service compensation does not create a bias against serving sick patients.
Quality measurement is thought to protect against distorted economic incentives by fostering competition based on improving quality. The theory is that measuring quality will give consumers a basis for choosing among health plans, and health plans a basis for choosing among providers, independent of cost. But this theory imposes unacceptable choices on consumers reasonably expect to avoid risky trade-offs between cost and quality in services like health care that involve their personal safety. Even to attempt such decision making is an emotional burden on patients, and then they still bear the risk of error and poor quality. This is especially true for Medicare patients, a disproportionate number of whom are emotionally, medically and economically ill-suited to make complex, uncertain, risky choices between competing health plans. Yet, the latest reform in Medicare mass-produces such a scheme for the Medicare population. (35)
Imposing such burden and risk on patients is especially difficult to justify when outcome comparisons are the measure of quality. Meaningful outcome comparisons require detailed risk-adjustment, which, as noted, is an uncertain endeavor at best. Moreover, even if adequate risk adjustment were achieved, there are a host of other practical and theoretical difficulties with the entire concept of choice based on outcome comparisons, whether the choice is made by consumers or health plans. (36) (37) In short, the prevailing consensus on the need for better outcome data to measure quality is misguided. Design flaws disable the health care system from acting on the quality information that already exists. Comparing outcomes to measure quality is a distraction from the fundamental need to achieve quality.
Key messages
• Tolerating even isolated failures of quality can increase costs exponentially.
• Failures of quality are not isolated but pervasive, because they are built into a system that permits undefined, inconsistent inputs and lack of patient involvement.
• The traditional analysis of competitive market theory in health care — that patients depend on expert physicians to exercise rational choice — must be reexamined.
• The reality is that not only consumers but producers of health care lack sufficient cognitive capacity to exercise rational choice effectively, unless the proper information tools are employed..
• Equipping patients to participate in medical decisions releases beneficial market forces, diminishing the need for third party micromanagement of medical decision making.