8. Deriving recommendations

When deriving recommendations a criterion is required which, when consistently applied, links the evidence to the recommendations. Guidelines normally contain many different recommendations based upon varying levels of evidence. For further adaptation of guidelines at lower levels of the health care system, it is important that local protocol developers and their users are aware of the type of evidence used to make the recommendations. It is advisable that explicit links between guideline recommendations and the supporting evidence are described to ensure user confidence in their validity²⁶.

The National Institutes of Health (NIH) of the United States of America have also developed guidelines for planning and managing consensus development conferences²⁷. This approach begins with a broad-based nongovernmental panel, which is selected to meet in public to review data on a specified topic. Presentations are made by invited experts and the panel concentrates on responding to questions about the topic. At the close of the meeting they come up with an agreed statement which is widely publicized.

A method of grading recommendations according to the evidence used has been proposed by the United States AHCPR (Table 3). For example, a recommendation based upon the results of at least one randomized controlled trial would thus be based upon level 1b evidence and would be termed a "Grade A recommendation".

Table 3: Grading recommendations

In the USPHS/IDSA HIV guidelines, a modified version of the categories suggested by PL Gross et al.²⁸ has been used to reflect the quality and strength of each recommendation.