Fragmented publishing: a large-scale study of health science

By Mette Brandt Eriksen, Tove Faber Frandsen, David Mortan Grøne Hammer & Janne Buck Christensen

The study, Fragmented publishing: a large-scale study of health science aimed at investigating, if there is a tendency of an increased number of publications per study in a large dataset from health science over time. The prevalence of fragmented publishing, where authors expand the number of publications by e.g. salami slicing, is a known phenomenon. Splitting data or results into several publications is in general considered a questionable research practice. There may, however, be several valid and reasonable arguments for doing so. The study is based on 6219 Cochrane reviews from the 53 Cochrane groups (9019 reviews were identified) published between 2012-2016. From these reviews, 92,428 studies and 139,779 publications were extracted and after filtering by inclusion criteria, a total of 55,181 studies and 86,533 publications formed the basis for the analysis. Results showed that most studies resulted in a single publication (in 95% of the cases a journal publication). In studies with several publications, the share of journal publications decreases, but still accounts for 4/5 of the study publication output. The development over time showed no overall significant mean number of publications per study, however there were a significant increase in the number of publications per study in a mid-period (1986-1999). Differences in mean number of publications per study were observed within 19 of the 53 Cochrane groups in the two stable time periods (1970-1985 vs. 2000-2015). Five of the 19 groups published less publications per study in the period 2000-2015 compared to 1970-1985 and 14 groups publishing more. There may be more reasonable reasons for an observed increased number of publications per study, e.g. larger studies. The majority (> 80%) of fragmented publications consists of journal publications. The extent of the fragmented publications identified in this study seemed to be tied to subdisciplines.  

Open Access fulltext (PDF)