Deconstructing the collaborative impact: Article and author characteristics that influence citation count

摘要

It is well known that collaborative papers tend to receivemore citations than solo-authored papers. Here we try toidentify the subtle factors of this collaborative effect byanalyzing metadata and citation counts for co-authoredpapers in the biomedical domain, after accounting forattributes known to be strong predictors of citation count.Article-level metadata were gathered from 98,000 PubMedarticle records categorized with the term breast neoplasm, atopic offering longevity and relevance across biomedicalsubdisciplines, and yielding a relatively large sample size.Open access citation data was obtained from PubMedCentral (PMC). Author-level attributes were encoded fromdisambiguated author name data in PubMed and appendedas article-level attributes of collaborations. A logisticregression model was built to assess the relative weights ofthese factors as influences on citation counts. As expected,the journal and language of the paper were the strongestpredictors. The significance of the number of authorsdiminished after accounting for other attributes. Some ofthe more subtle predictors included the group’s highest hindex,which was positively correlated, while the diversityof author h-indices, minimum professional age, andauthor’s total unique collaborators were negativelycorrelated. These observations indicate that smallercollaborations composed of early superstars – young,rapidly successful researchers with relatively high andsimilar h-indices – may be at least as influential inbiomedical research as larger collaborations with differentdemographics. While minimum h-index was important, thefirst author’s h-index was insignificant, underscoring theimportance of the middle authors’ publishing history. Thegender diversity outcomes suggest that mixed groups maybe ideal, and further research in this area is indicated.