[Criteria for defining consensus achievement in Delphi studies that assess potentially inappropriate medications in the elderly].

Background: The statistical analysis of survey responses based on a categorical Likert scale, as those used in the Delphi technique, has been questioned because the methodology used is based on numerical data analysis.

Aim: To develop criteria for defining the consensus achievement in Delphi studies conducted for the assessment of potentially inappropriate medications (PIM) in the elderly.

Methods: It was constructed an index (Yq) which takes into account the agreement by pairs of evaluators, whose calculus equation is based on combinatorial analysis. Yq was applied in a pilot study with Delphi methodology for assessing the safety of 12 drugs through a Likert scale with five response categories.

Results: On the basis of analyzing the distance (d) in between each pair of categories from the scale, it was determined the associated weighting, w=1-d, to be applied in the calculus of Yq: proportion of agreements weighted. There were defined three criteria that need to be satisfied to obtained the consensus in each item (drug) of Delphi questionnaire : a) Number of evaluators ≥60% of the panel members, b)Yq≥0,800; c) frequency of the statistical mode ≥60%. On regards to the analysis, 8/12 drugs were evaluated as potentially inappropriate for elderly adults while the remaining 4 should be revalued in successive rounds due to not obtaining consensus.

Conclusion: The index takes into account the real distances between the Likert scale categories and the developed criteria constitute a simple tool for the analysis of the Delphi questionnaires made for the valuation of PIM on older adults.