A few panel members dominated guideline development meeting discussions: Social network analysis.

OBJECTIVES: To identify patterns of interactions that may influence guideline panels' decision-making. STUDY DESIGN AND SETTING: Social network analysis (SNA) to describe the conversation network in a guideline development meeting in United States. RESULTS: We analyzed one two-day guideline panel meeting that included 20 members who developed a guideline using the GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach. The conversation pattern of the guideline panel indicated a well-connected network (density=0.59, clustering coefficient=0.82). GRADE topics on quality of evidence and benefits versus harms accounted for 46%; non-GRADE factors accounted for 30% of discussion. The chair, co-chair and methodologist initiated 53% and received 60% of all communications in the meeting; 42% of their communications occurred among themselves. SNA metrics (eigenvector, betweenness and closeness) indicated that these individuals also exerted highest influence on discussion, controlled information flow and were at the center of all communications. Members were more likely to continue previous discussion with the same individuals after both morning breaks (r=0.54, P<0.005; r=0.17, P=0.04), and after the last break on day 2 (r=0.44, P=0.015). CONCLUSION: Non-GRADE factors such as breaks, and the members' roles, affect guideline development more than previously recognized. Collectively, the chair, co-chair and methodologist dominated the discussion.