Allocation Versus Evidence-Based Medicine

Decisions based solely on evidence of the efficacy of medical care are not rationing decisions. There is no medical obligation to provide and no societal obligation to pay for care that is harmful or ineffective. In fact, clinicians use special terms to describe interventions that fall into these categories, including “futile,” “not standard of care,” “medically inappropriate,” “wasteful,” or “experimental.”^5,6 For example, an intensivist who decides not to transfuse a critically ill patient with a hematocrit of 27 is not rationing blood, even though blood is an expensive and limited resource; in this case, there is evidence that a transfusion would be of no benefit and might even be harmful.⁷ Likewise, the decision not to use human growth hormone, an expensive medication, in a chronically critically ill patient is not a rationing decision, because this treatment has been shown to be ineffective and may be harmful.⁸

Unfortunately, assessments of benefit and harm are not as straightforward as the terms would suggest, and the line between effective, ineffective, and experimental treatment often is a personal decision for the individual clinician. Decision science has taught us that medical decision making is a complex process that frequently obscures the true rationale for the choice.⁹ In fact, judgments allegedly based solely on objective evidence of safety and benefit often incorporate a variety of subjective values and biases.¹⁰ These may include the value the clinician assigns to being wrong; the value assigned to trying to “rescue” a patient in imminent danger of death; the clinician’s tolerance for uncertainty; the impact of the decision on the clinician’s finances; biases about the patient’s race, gender, functional status, or age; and the cost or availability of the resource.¹¹ The transition from statements that summarize the evidence of benefit to recommendations that incorporate cost and other values is often very subtle. For example, the authors of a recent systematic review of colloid resuscitation in critical care conclude that “there is no evidence from randomized controlled trials that resuscitation with colloids reduces the risk of death compared to crystalloids in patients with trauma, burns and following surgery.”¹² This is a statement of their summary of the evidence of efficacy of colloid therapy. Like many treatments in critical care, the evidence neither supports nor completely refutes the use of colloids as resuscitation fluids in the critically ill. However, the authors conclude, “As colloids are not associated with an improvement in survival, and as they are more expensive than crystalloids, it is hard to see how their continued use in these patient types can be justified outside the context of randomized controlled trials.” Whereas the first statement may be a fair summary of the evidence, the recommendation against using colloids in the second sentence is not based solely on the evidence. It incorporates an implicit rationing strategy that pays only for treatments that have demonstrated benefit in a certain way. Although one might conclude from the authors’ review that colloid resuscitation is experimental or that its benefit is likely to be small, the reasoning for recommending against its use is based on the cost of the treatment. Presumably, if colloid fluids were the same price as crystalloids, the authors might reach different conclusions, even though the cost does not change the evidence of efficacy.

The preceding example shows how assessments of cost can creep into recommendations for therapy even without a formal discussion of allocation. Because clinicians and payers may be reluctant to admit they are incorporating cost or availability into the rationale for a decision, they may find decisions based on futility or appropriateness less ethically problematic than those based on rationing. In fact, these judgments may implicitly contain assessments of cost by incorporating cost into the definition. When is there sufficient evidence to move a treatment or diagnostic device from experimental care to standard care? When is there sufficient evidence, absent evidence of outright harm, that a treatment is ineffective as opposed to not yet of proven efficacy? These decisions are frequently made by consensus bodies using subjective or poorly characterized criteria. The evidence threshold tends to be higher for treatments that are risky or expensive or for which there is no alternative. Conversely, the threshold for accepting a treatment as “standard” is lower if that treatment is inexpensive and safe and offers the potential of rescuing a patient in imminent danger of dying. For example, consider the decision to elevate the head of the bed of mechanically ventilated patients to prevent ventilator-associated pneumonia. This is an inexpensive and safe treatment to offer patients. It might take less evidence to convince clinicians to use this treatment than to use kinetic beds or topical prophylactic antibiotics, which are more expensive and may raise safety issues. Therefore, the cost of an intervention may be incorporated into the assessment of whether it is a standard of care.

These judgments are further complicated by the motivation of the decision maker. It would be difficult for an insurance company that is assessing whether a specific therapy is experimental or standard of care to be unbiased, because its decision will affect its profits. Alternatively, surgeons who developed a procedure may be committed to its benefits in a way that compromises an objective evaluation. The complexity of the assessment of efficacy and cost highlights the importance of making allocation decisions as objective, explicit, and public as possible. Because medical decisions are so complex, and because decisions in the ICU are further complicated by their immediacy and the severity of patients’ illnesses, it is essential that clinicians understand their own motivations and the evidence supporting their decisions and have a process in place for allocating resources.

Allocation Strategies

Allocation decisions are usually separated into macro-allocation decisions (involving groups of people and usually made at a managerial or health policy level) and micro-allocation decisions (made at the bedside and involving specific cases). A hospital’s decision not to hire additional ICU nurses is a macro-allocation decision; a nurse-manager’s decision to allocate a specific patient to share a nurse in the ICU rather than to receive 1 : 1 nursing is a micro-allocation decision. This chapter is concerned primarily with bedside, or micro-allocation, decisions that clinicians make on a routine basis. There is an important interaction between micro- and macro-allocation decisions, because macro-allocation decisions ultimately affect individuals, and macro-allocation regulations are an effective rationing strategy (Table 215-2). There are a number of approaches to allocating resources (see Table 215-1). Although they are all feasible, they are not all equally ethical.

TABLE 215-2 Allocation Decisions at Different Levels

The principles of equality, fairness, justice, and due process make some strategies less acceptable. The principle of utilitarianism directs resource allocation to maximize the “utility” or benefit to the greatest number of people for any given amount of resources. To the extent that utility can be determined by measuring patient outcomes such as health-related quality of life, and to the extent that we can theoretically estimate the effects of medical treatments on utility, we can calculate exactly which set of medical treatments to pay for to maximize the benefit to the population. These studies are called cost-effectiveness analyses and are the quantitative embodiment of utilitarianism.

Allocating medical resources through cost-effectiveness analysis has important limitations. First, medical cost-effectiveness analysis cannot tell how much money to allocate to medical care as opposed to other goods and services; it can only determine how to maximize health outcomes for a selected outlay of resources. Second, cost-effectiveness analysis may not fully account for some factors society values. For example, cost-effectiveness analysis routinely treats all human lives as equally valuable; however, society often places a high value on saving identifiable lives in imminent danger of death, and it may not value additional years of life in the elderly as highly as additional years of life in the young.¹³ Cost-effectiveness and other utility-based allocation strategies fail to account for the value society places on rescuing lives in imminent peril—a not uncommon occurrence in the ICU.¹⁴ Standard economic analyses may not value equal distribution as much as optimal distribution and, to this end, may discriminate in settings society finds unacceptable.¹⁵ Finally, cost-effectiveness analysis is a mathematical technique that generates comparative outcomes for populations of patients. It is meaningless to speak of a treatment as being “cost-effective” for an individual.

The primary value of cost-effectiveness analysis as an allocation tool is the ability to compare various strategies.¹⁶ For example, one can compare the cost-effectiveness of captopril versus no captopril in survivors of myocardial infarction with the use of fluoxetine versus imipramine for major depression to decide whether to use captopril, fluoxetine, both, or neither. Cost-effectiveness analysis provides a ruler, in terms of dollars per life-year or dollars per quality-adjusted life-year (QALY), that allows different treatments for different diseases to be compared. The crucial data that must be available to make these comparisons is information on the treatments’ effects on survival or health-related quality of life. Unfortunately, in critical care, the number of treatments shown to improve survival or health-related quality of life is small. Although we have data on strategies to reduce gastrointestinal bleeding, duration of mechanical ventilation, and catheter-related infections, none of these interventions has been shown to affect QALYs.^17–19 Therefore, the cost-effectiveness analyses for these interventions are expressed as, for example, dollars per gastrointestinal bleed prevented.²⁰ These ratios cannot be used to compare a treatment to prevent gastrointestinal bleeding with a treatment for myocardial infarction, because the latter is expressed in dollars per QALY. Cost-effectiveness analyses with non-QALY denominators can be helpful in bedside rationing decisions when the intervention is shown to be equally or more effective and reduces cost. For example, special beds in the ICU both prevent decubitus ulcers and reduce the overall cost of care, even when the cost of the bed is factored in. Therefore, the cost-effectiveness ratio (expressed in dollars per decubitus ulcer prevented) is a negative number.²¹

Illusory Cost Savings

Since the earliest days of intensive care, technologic, workforce, and organizational innovations have been proposed as opportunities to reduce the exorbitant cost of critical care. In 1973, an optimistic author wrote, “[the] more promising approaches to cost reduction are all in an early stage of development now. Both deprofessionalization of the ICU by wider use of allied health personnel, and the automation of therapeutic functions are just beginning to be applied.”²² Despite the implementation of both these measures, there is little evidence that cost increases in hospital or ICU care have been curbed by technologic innovation. In fact, the opposite has occurred. This is not surprising, because technologic innovation in other areas of health care, though often associated with better outcomes, is rarely a source of cost savings.

Cost analyses are problematic in medical care, and critical investigators must be able to identify cost savings that are real and that will appear in their budgets from savings in indirect costs that will be accrued elsewhere.²³ There are several common but problematic arguments about cost reduction in critical care: (1) that reduced ICU length of stay will reduce the cost of care in the ICU, (2) that ordering fewer tests will reduce the cost of care in the ICU, and (3) that fewer admissions of futile-care patients will save money. It is important to recognize that not all calculated cost savings will be realized at the ICU or hospital level.

ICU costs are frequently inferred from length of stay. For example, in a cost-effectiveness analysis of antibiotic-coated catheters, the authors assigned a cost of $9738 to a catheter-related bloodstream infection.²⁴ Epidemiologic studies show that patients with catheter-related infections spend more time in the hospital, even after controlling for severity of illness.²⁵ The cost of a catheter-related infection is, in part, derived by simply multiplying the estimated number of extra days spent in the hospital by the cost (based on hospital charges) of a day in the ICU or ward. In fact, we do not really know whether using antibiotic-coated catheters shortens ICU length of stay, because the randomized trials demonstrating that they prevent infection were not sufficiently powered or did not show a reduction in mortality or length of stay.¹⁹ Even if antibiotic-coated catheters do reduce length of stay, money “saved” by reducing length of stay is a different kind of money from that used to buy the catheters. By reducing length of stay, the ICU will be able to care for more patients, but they will be sicker and more expensive patients.

Identifying treatments for specific conditions in the ICU that reduce overall costs, even if they have no effect on QALYs, is extremely useful to the intensivist who must allocate resources. Implementing economically dominant strategies is an easy allocation decision, because they reduce costs but do not worsen patient outcomes. However, predicting the actual effect of any decision on actual costs in an ICU or hospital is complex because each hospital performs cost accounting and budgeting in idiosyncratic ways.

The effect of different payer mixes, contracts for nursing and respiratory therapist labor, allocation of indirect costs, and whether the ICU budget is fixed or grows with the number of patients served all influence whether allocation decisions accrue savings that can be appreciated at the ICU level. For example, the drug acquisition costs of once-daily medications are frequently higher than the costs of medications given more frequently. However, there are labor costs associated with administering medication more frequently that may offset the costs of the once-daily medication. Unfortunately, unless changing to once-daily medication reduces the workload to the point where it is feasible to actually reduce the number of nurses, there will be little realized savings. This is because labor costs are not infinitely scaleable. Even if there is 15% less work to do, it may not be possible to hire 15% fewer nursing hours. Patients who need 1 : 1 nursing care will continue to need this level of care regardless of whether the nurses are administering once-daily medication or not. It may be that changing medication routines improves care by using nursing time more efficiently, but this may not be reflected in a cost reduction. A reasonable criterion to consider for a proposed cost-saving intervention is whether it will reduce the number of staff that need to be hired or whether it can reduce acquisition costs for equipment or medications. If it will not, then cost savings are not likely to be realized in the ICU.

The cost estimate used in many cost-effectiveness analyses assumes that every day in the ICU costs the same. This is certainly true for what the hospital charges, but it is not true in reality. The first few days in the hospital and ICU are generally far more expensive than the last days.²⁶ Patients are more likely to require active interventions and closer nursing care in the early days in the ICU. Clearly, interventions that reduce ICU length of stay cannot reduce early days in the ICU; they simply eliminate later lower-cost days. This is rarely accounted for in cost analyses. This was validated at the national level as U.S. healthcare costs peaked during a period when hospital inpatient days declined by 40%.²⁷ Therefore, standard cost analyses overestimate cost savings likely to be realized by reducing length of stay.

Reducing test ordering in the ICU has been offered as a technique for cost reduction. This too is a perfectly reasonable option on clinical grounds. Overtesting yields increased false-positive results, which may lead to clinical complications in search of diseases that never existed. However, the actual cost reduction at the ICU level achieved by limiting test ordering is likely to be overestimated in a simple charge-based analysis. The actual marginal cost of performing the 101st arterial blood gas once the analyzer has been purchased and the technician has been paid to perform 100 arterial blood gases is minimal. If reductions in test ordering are of sufficient magnitude to staff the laboratory with fewer people or to forgo purchasing new equipment, significant cost reductions can be realized. In fact, depending on how indirect costs in the hospital are allocated, it is possible a reduction in test ordering will place the clinical laboratory under considerable budgetary constraints. Fewer tests may reduce the amount of money the laboratory director receives to cover staff costs, which may not decrease in the same proportion as test ordering.

Patients may be admitted to the ICU even when they have a negligible chance of survival. It seems reasonable to assume that if these patients receive care outside the ICU, resources that would have been expended without benefit in the ICU will be saved. On its face, this appears to be the sort of painless cost saving intensivists should look for. Unfortunately, a careful analysis of potential savings from limiting care at the end of life shows that such care accounts for a relatively small amount of overall healthcare spending, that implementing these strategies may worsen overall health outcomes by affecting the care nonterminal patients receive, and that care would have to be withheld from young patients (some of whom would have had prolonged survival) to achieve any savings.²⁸

Strategies for Bedside Allocation of Resources in the Intensive Care Unit

Ultimately, allocation decisions occur at the bedside in the ICU. A number of studies demonstrate that under settings of restricted access to ICU beds, physicians allocate these beds on the basis of severity of illness. In these situations, the average severity of illness in the ICU increases, as it does on the hospital ward.²⁹ Unfortunately, these decisions are also driven by arbitrary factors including patient age and gender, reimbursement, and physician power in the institution.³⁰ It is important that clinicians plan in advance for such difficult decisions so their deliberations are explicit, open, and guided by principles rather than ad hoc case-by-case decisions.

Conclusion

Allocation of resources in medicine is an unavoidable process. Clinicians do have control over whether these decisions are implicit or explicit, whether they are made after open discourse or with no discussion, and whether the decisions are informed by the available literature. Clinicians in the ICU may in fact face fewer implicit allocation decisions than their colleagues in other areas because of the imminent risk of death in the ICU and the value society places on protecting those lives. In fact, there is relatively little empirical evidence of how often intensive care services are allocated. The effect of different interventions on actual costs varies depending on local factors such as reimbursement and indirect cost allocation. Allocating ICU beds is the most challenging allocation decision most intensivists will face. The best time to handle these situations is before they occur. Public, explicit triage and discharge criteria that are developed in collaboration with ICU users (emergency department, surgery, oncology) well in advance of the actual decisions are essential for fair and efficient use of intensive care resources.

Key Points