Editorials

Guidelines for clinical guidelines

A simple, pragmatic strategy for guideline development 

Clinical guidelines are systematically developed statements designed to help practitioners and patients make decisions about appropriate health care for specific circumstances.¹ Clinicians are being inundated by a tidal wave of guidelines. In Britain alone, regional programmes for audit have recently developed about 2000 guidelines or protocols.² In addition to numerous clinical guidelines, a number of "guidelines for guidelines" have been produced, ranging from simple³ to complex.⁴ These reflect the increasing attention being paid to the methodology of guidelines development and the validity of guideline recommendations. While we support the increasingly rigorous approach taken to guideline development, it is important to re-emphasise the central role of guidelines themselves, which is to help clinicians make better decisions.

Clinicians need simple, patient specific, user friendly guidelines. We highlight three key components for such guidelines. The first is the explicit identification of the major decisions, relevant to patients, which have to be made, and the possible consequences of these decisions. Many encounters with patients involve multiple decisions, so the key to developing usable guidelines is to identify only the most important ones. These decisions and their consequences may often be difficult to map, and remarkably little is known about how doctors actually make decisions.⁵ But unless we limit guidelines to the major decision points, they are likely to be too unwieldy to use in practice.

The key clinical decisions generally relate to making a diagnosis; estimating prognosis; assessing relevant outcomes, including the benefits, risks, and costs of alternative treatments; and, finally, weighing up the various consequences of different treatment options. It should be possible to produce a flow diagram or algorithm which identifies the key decisions and important outcomes relevant to patients and others.

The second component of successful guidelines involves bringing together the relevant, valid evidence that clinicians need to make informed decisions at each of the key decision points. This remains the most difficult step as, in many if not most clinical areas, the necessary research evidence is inadequate, and methodologists continue to struggle with these shortcomings. The increasing interest in evidence based practice and guidelines has highlighted the huge gaps in the evidence, although recent studies suggest that the potential to make evidence based decisions may be greater than generally believed.⁶ Moreover, groups such as the Cochrane Collaboration, the York Centre for Reviews and Dissemination, and the United States Agency for Health Care Policy and Research, will increasingly be able to provide guideline developers with comprehensive and systematic overviews of the evidence.

One of the cornerstones of evidence based practice (and evidence based clinical guidelines) is the requirement that the evidence is relevant to individual patients.⁷ Much of the evidence presented in formal overviews, although comprehensive and valid, is not in a form directly relevant to individual patient care. For example, overviews typically summarise treatment effects in terms of relative risks or benefits, whereas treatment decisions, where possible, should be guided by the absolute risks and benefits of treatment. These measures can be presented in units such as events per 100 patients treated (or untreated) per year, or the number of patients who would need to be treated to prevent an event (number needed to treat).⁸ An example can be viewed at the following website: http://cebm.jr2.ox.ac.uk/docs/prognosis.html. As more research on the cost-effectiveness of treatments is published, this too needs to be incorporated in a relevant form into guidelines. The more guidelines can present evidence which can be applied to individual patients, the more useful they will be for real life clinical decision making. Explicit statements about the benefits and risks of treatment can then be weighed by patient preferences and available resources.⁹ This is currently difficult to achieve for most clinical problems, making it necessary to write more general guidelines,¹⁰ which are less explicit because the evidence does not allow the calculation of outcome probabilities. None the less, guideline developers should be encouraged to follow the process outlined above and acknowledge where recommendations are based on inadequate evidence.

A third essential component of a successful guideline is the presentation of evidence and recommendations in a concise, accessible format. Decision makers must be able to retrieve and assimilate information quickly. Moreover, information must be presented in a flexible format that is applicable to specific patients or circumstances. As clinicians move into the computer age, the possibilities of more immediate access to the relevant evidence will increase.

We consider these three components to be basic building blocks of usable clinical guidelines. The embarrassingly wide variation in much healthcare practice suggests clinicians use different information to inform the same decisions. In a significant proportion of clinical situations, guidelines could become the common language enabling patients, practitioners, scientists, and purchasers the opportunity to share information more effectively. However, "guidelines for clinical guidelines" need to be kept simple and need to focus on the essential components of usable guidelines. Unless we can communicate a simple, pragmatic strategy for guideline development, we will continue to be embarrassed by variations in clinical guidelines as we are by inappropriate variation in clinical practice.