There are multiple “evidence-based” diabetes guidelines that are used to develop diabetes performance measures for diabetes by multiple stakeholders within the United States, Europe, and other countries.12 Similarly, there are multiple differing sets of performance measures. Although leading organisations that develop measures or endorse measures report their process, the explicit reasoning underlying the selection of the final measures is often not apparent.3 In the current issue of QSHC, Martirosyan et al report on their use of the RAND Panel Delphi methodology to develop valid prescribing quality indicators (PQIs) for persons with type 2 diabetes (see page 318).4 This is a good step forward, although it is not without shortcomings. In their report, a panel of experts was given practice guidelines and the evidence base. A series of proposed PQIs were then evaluated in a two-round process. Their paper raises several issues about performance measures and how they are chosen.

DIFFERENT PEOPLE LOOK AT THE SAME EVIDENCE AND CONCLUDE DIFFERENT THINGS

Experts clearly disagree on how to evaluate evidence, and the Delphi method is designed to achieve consensus during the decision-making process. The Delphi method provides a structure to organise group communication. It is an iterative process of repeated evaluations by a group of individuals whose opinions or judgements are of interest. It is designed to secure the benefits of group decision-making while insulating the process from its limitations, for example, dominant individual members, and is useful for situations when the decisive factors are subjective and not knowledge-based, that is, where individual judgements vary and there is incomplete knowledge. It is usually conducted as a series of sequential questionnaires or “rounds.” After the first round, each subsequent round is accompanied by anonymously presented information regarding the preceding round. Each panel member is thus encouraged to reconsider and, if appropriate, to change their previous reply in light …