PRM223 - Social Network Analysis of Authorship Networks and the Identification of Expert Advisors

Abstract

Systematic reviews are often supplemented with the use of external experts to provide guidance on the nuances of the area. This can help add context if a review is used to support trial design or health economic model development. The ideal expert would have a deep understanding of the area and be well connected to those individuals conducting trials. The aim of the current research was to assess whether social network analysis of coauthor networks could be used to rapidly and objectively identify individuals with the qualities desired in an external expert. Publication lists from a recent systematic review of rheumatoid arthritis were used to produce a list of links between authors and publications. This was then imported into the Gephi program for social network analysis. Within Gephi, matrix multiplication was used to transform this network into a coauthorship network. Eigenvector centrality was then used to infer the amount of access individual authors have to the research community as a whole. The use of eigenvector centrality as a measure of influence within the author network was then validated by correlating the centrality scores of a random sample of authors against independent ratings of desirability of those individuals’ expertise. The coauthor network for rheumatoid arthritis, while not completely connected, showed a high degree of connectivity (mean degree: 26, network diameter: 5). Eigenvector centrality allowed the identification of key experts, with the highest scoring experts each providing direct access to approximately half of the whole network. Eigenvector centrality measures were a reliable predictor of mean desirability scores from ten raters (F(1,9)=20.35, p=0.0015, R-squared=0.69). Social network analysis of coauthor networks provides an efficient and robust method for the identification of expertise, and can be used as part of the systematic review process.