Cookbook medicine.

As medical director at an NCQA-accredited health maintenance organization, one of my tasks is to develop and implement practice guidelines in my organization. Because my company contracts with large medical groups that contract with many other HMOs, implementing practice guidelines we develop is very difficult. A physician cannot keep 17 guideline books on a shelf and refer to the manual of each patient's HMO before deciding how to treat the patient.

I discussed this challenge with a friend whose only contact with the health care system had been as a patient. Beth (a pseudonym) was at first repelled by the thought of guidelines. She wanted her physicians to tailor diagnosis and treatment to her, not follow some standard protocol developed for the average person. Like most people, she does not consider herself average.

Beth agreed, of course, that she wanted treatments that resulted in the best outcomes the current state of medical science could predict. She recognized that some physicians might be out of date or make mistakes, and she could see the need for some safeguards against that possibility. She could see why I might want to require some physicians to perform at some minimum level. She believed her physicians routinely performed above that minimum level, however. She resented any outside interference that might deter them from providing her with the best treatment tailored just for her.

I explained why practice guidelines are called "guidelines" and not "requirements": the fear that developing anything more rigorous than guides for treatment will result in "cookbook medicine." I searched for a metaphor to explain what I hoped to accomplish. Beth had worked in a restaurant in the past, and the metaphor became clear to me. Cookbook medicine is exactly what we need.

The best chefs may, through trial and error, create marvelous meals without recipes. But many creations will be flops. We have cookbooks, in part, because most chefs do not have the time or resources to experiment. They want the best consistent outcomes. Recipes may stifle some creativity, but they will give an acceptable result most of the time, if they are good.

The worst. that happens in cooking, if an experiment goes bad, is the waste of a few ingredients. In medicine, a failed experiment is a bad outcome. Most patients will choose a proven recipe over risking an experiment that could result in a better outcome but is more likely to result in a worse one.

Recipes are most important in higher volume restaurants, where meals must be mass produced at a consistent level of quality. The chef is captain of a team. Each member of the cooking team must know his or her role in the overall process. The chef may oversee the process. He or she might decide to make small variations based on varying inputs. The chef might find one ingredient is of insufficient quality and therefore substitute another. Perhaps the baby potatoes the chef usually uses cannot be found, large potatoes are bought instead. The chef instructs the vegetable cutter how to cut up the larger potato and adjusts the cooking time. In other words, the chef makes adjustments but ensures that all other members of the team understand the reason for the adjustment.

There is room for artistry and judgment, in cooking and in medicine. Artistry alone does not create enough meals to feed every patron. Recipes are a means of multiplying that artistry for greater effect. It allows the chef to be the captain of a team, rather than requiring him or her to perform every aspect of cooking. Most people cannot afford a private chef. Most people cannot afford personal physicians. If we are to provide for the entire population, we must use a team, and a team needs a recipe or a playbook. Sometimes we do not know which recipe is best. We must throw away the cookbook for research and try different approaches. But when a good recipe exists, we cannot risk a bad outcome by experimenting on a patient. Even research in that situation has ethical pitfalls.

Practice guidelines are becoming an integral part of medical management. Large employers sponsor them and demand them of their health care suppliers.[1,2] The National Committee for Quality Assurance has made a process for developing, approving, and implementing guidelines a requirement of accreditation. As a result, many HMOs spend resources developing their own internal guidelines.[3] Others collect the many guidelines that are written by professional societies and government agencies and try to chose one for their patient populations. Doing so generates a great deal of confusion.

At its best, a practice guideline is an analysis of what works in medicine and what does not. National experts analyze all the data available to determine the best way to diagnose a condition, to evaluate a complaint, or to undertake a procedure. Which tests truly add-to the diagnostic process? Which medications truly improve outcome?

Creating consensus panel guidelines is an expensive and time-consuming process. The medical literature must be searched. All relevant articles must be compared in methodology as well as results. Medical experts and statisticians are paid to perform these comparisons. Meetings must be convened. Because the data are rarely clear-cut, differences of opinion exist. Compromises must be negotiated. Finally, a guideline is produced and published. Medical research continues, however. New technology replaces the old. The guideline rapidly falls out of date. Thus, the development expense is not a long-term investment.

Most health care delivery organizations do not have the resources to develop consensus guidelines. The best they can do is subscribe to information from expert panels. Their main role is to communicate the latest best assessments to practicing physicians. Communication of this sort is not a new challenge. The purpose of textbooks and clinical journals is to disseminate new data to those who will use the information in the practice of medicine.

The amount of data now available overwhelms practicing clinicians. Many physicians have lost their skills in analyzing data. They rely on consultants, advertisements, drug detailers, and other less rigorous methods of incorporating new knowledge. Recent advances in decision analysis, meta-analyses, and other techniques have made the task of analyzing data more specialized. Clinicians may be wise to allow panels of experts to analyze the data and make recommendations. Where clear benefit can be demonstrated for one course over another, physicians can follow the appropriate guidelines.

One might think that the editors of textbooks fill this need. Many textbook chapters are written by one author, however, and are weighted by that physician's clinical experience. Expert panels incorporate a greater breadth of clinical experience.[4] Many textbooks emphasize pathophysiology and are not in the brief, algorithm format of good guidelines. Worse, textbooks are often out of date by the time they are printed. Yet, some texts are updated frequently; are based on meta-analyses; and are reviewed, if not written, by several experts. Such texts ought to fill the needs for authoritative treatment recommendations.

Each specialty college or society could publish a loose-leaf textbook of diagnostic and treatment protocols for various complaints and illnesses. Panels of experts on the disease in question would write chapters and update them frequently. This hypothetical document would be much more useful to physicians than most current guidelines.

If a guideline is merely a chapter of a textbook, an educational source for busy practicing clinicians, why do employers and regulatory agencies require them? The answer, of course, is that they are not intended as educational tools. The purpose of guidelines, as customers define them, is to change physician behavior. Studies have shown the limited success of education in altering physician performance.[5]

Guidelines can affect practice by three mechanisms. They can be used to determine insurance coverage decisions.[2,6-8] They can become utilization review protocols to preauthorize care or to deny authorization for care outside the guideline. Or they could be used as quality assurance indicators or audit tools. Administrative systems to affect resource use are effective.[5] Both insurance coverage and prior authorization systems fall into this category.

Using guidelines as quality assurance indicators requires some basic rethinking about the structure of guidelines. More and more, quality is being defined through continuous quality improvement (CQI) terminology as conformance to requirements. Guidelines are either requirements for behavior or they are not. If they are only educational, they are not requirements and cannot be used to measure quality. Failure to conform to a nonrequirement is not a quality problem.

If we can determine that one treatment is better than another, we should require that physicians follow the better treatment course. Failure to do so jeopardizes patients and quality. If a treatment is better in some circumstances but not others; we should change the scope of the requirement. We limit the requirement, and therefore the guideline, to the subset for which benefit is proven. We should not change the requirement itself. Requirements are defined by outcomes. As long as following a guideline results in improved outcome, we should retain the requirements within the guideline.

The concept of quality as conformance to requirements allows us to measure quality. We should not dilute the meaning of requirements by making them optional, or variable, or based on conjecture. Defining requirements and using them as mandatory standards against which physician performance is judged can improve care. Once we measure performance, we can improve it. A concerted national effort to define and disseminate the processes that create the best outcome could also improve care.

Unfortunately, the current guideline process is haphazard. The American Medical Association's laudable attempt to collect 1,600 guidelines[9] reveals that many are individually published pamphlets or monographs. Some are brief, some are many pages long, some are textbooks. Some are algorithms, others prose, others in outline form. This hodgepodge is not very useful to a practicing clinician. In fact, they are not very useful to a medical manager.

I believe that guidelines should be used as QA and UR tools. They should be defined as requirements. Their scope should be as narrow as necessary to make them valid requirements for processes that truly affect patient outcomes. If they are merely educational tools, they will be less useful than textbooks of medicine, which at least are indexed and in an accessible format.

Local guidelines can be useful as recipes for care within one organization. Should primary care physicians get an IVP before or after referral to the urologist? What physical modalities should be used in managing back pain before referring to orthopedics? If these rules are clear, patients will get to the right specialist expeditiously, without wasted referrals and tests. Quality will improve because of reduced process times, and costs will fall. Because such rules will depend on the expertise of the individuals within a group, they should be developed locally.

Managed care organizations or individual hospitals can and should develop guidelines to increase consistency, reduce redundancy and counterproductive behaviors, and reduce variations in processes. These organizations should not develop "consensus" guidelines. A driving purpose of the guidelines movement is to reduce geographic variation in practice. Locally developed guidelines only foster more such variation. Managed care organizations and hospitals can disseminate and implement national consensus guidelines and can measure conformance to them.

The term "guidelines" is itself part of the problem in implementation. The word was adopted in the hope that it is less threatening to physicians than are the terms "standards" or "requirements." Yet many physicians still fear that guidelines will be used against them, to limit their autonomy, to create standardized treatment at the patient's expense. And they are correct. The authors of a guideline can be vague about the patient subsets it applies to, and whether any exceptions apply, because it is just a guideline. The main customers of the AMA Practice Guideline Compendium are plaintiff attorneys, who will use them as evidence of the standard of care, and health administrators, who will use them for coverage decisions. Vague guidelines do not benefit physicians, who can just as well look up treatment in a good textbook. Physicians can use requirements. Expert panels can determine the best recipe for specific patient subsets for specific diseases or treatments of disease.

Physicians can then follow the recipe. For many, perhaps most, patient subsets, data are inadequate to determine the best course. There may be as few as a couple of dozen diseases for which requirements can be established. That is enough to start. The term "requirements" seems more limiting, harsher, than "guidelines." But let us call them what they are. Quality improvement efforts to reduce variation and optimize outcome will not work unless we treat them as requirements.

By clearly labeling them as requirements, their developers will be forced to be very specific about the subsets of patients to which they apply. For example, instead of a guideline for the management of gallbladder disease, we may develop a requirement set for the patient between 20 and 50 years old, with symptomatic, radiologically proven cholelithiasis and with no co-existing disease. That requirement set may include the additional history and physical findings that should be documented, the laboratory tests that should be performed, and perhaps an algorithm for treatment based on the results. We can use such documents as algorithms for nonphysician health practitioners, as QA and UR tools, and as educational tools for physicians.

Eventually, once physicians determine the best recipe for a patient subset, physician assistants or nurse practitioners can manage patients in that subset. The physician's role will be to develop the recipe. Physicians will determine when a patient falls into a subset for which there is a guideline and will manage patients who do not fall into a subset. For the latter patients, clinical judgment is needed to fill the gaps in medical knowledge. This is the art of medicine. Requirement sets will allow us to focus our improvement efforts on those diseases for which data are adequate and to improve quality where we can define and measure it. We can use outcomes data to refine these requirements to meet our objective of the optimal health of our patients.

References

[1.] Nash, D., and Markson, L. "Managing Outcomes: The Perspective of the Players." Frontiers of Health Services Management 8(2):3-51, Winter 1991. [2.] Hadorn, D. "Defining Basic Health Care Benefits Using Clinical Guidelines." Prepared for "Creating a Fair and Reasonable Basic Benefit Plan Using Clinical Guidelines," sponsored by the California Public Employees Retirement System's Health Benefits Advisory Council, Sacramento, Calif., April 1991. [3.] Gottlieb, L., and others. "Clinical Practice Guidelines at an HMO: Development and Implementation in a Quality Improvement Model." QRB 16(2):80-6, February 1990. [4.] Sox, H., and Griner, P. "An Invitation to Join a Controversy." Annals of Internal Medicine 116(5):422-423, March 1, 1992. [5.] Eisenberg, J. Doctors' Decision Making and the Cost of Medical Care. Ann Arbor, Mich.: Health Administration Press, 1986. [6.] Hadorn, D., and Brook, R. "The Health Care Resource Allocation Debate: Defining Our Terms." Prepared for "Creating a Fair and Reasonable Basic Benefit Plan Using Clinical Guidelines," sponsored by the California Public Employees Retirement System's Health Benefits Advisory Council, March 1991. [7.] Brook, R. "Health, Health Insurance, and the Uninsured." JAMA 265(21):2998-3002, May, 29, 1991. [8.] Eddy, D. "What Care Is |Essential?' What Services Are |Basic?'" JAMA 265(6):782,786-8, Feb. 13, 1991. [9.] Directory of Practice Parameters. Chicago, Ill.: American Medical Association, 1990.