CHANGES DURING PEER REVIEW:
CHARACTERIZING THE EVOLUTION OF A CLINICAL PAPER

Gretchen P Purcell1 and Frank Davidoff
Department of Surgery, Duke Hospital North, Durham, NC 27710; Annals of Internal Medicine, Philadelphia, PA, USA

Objective: We sought to categorize the inadequacies in reports of clinical research and to characterize quantitatively the resultant changes during the peer-review and editorial processes.

Design: We selected original research articles from 2 issues of Annals of Internal Medicine and identified all substantive textual changes between the submitted manuscripts and the corresponding published papers. We characterized the reasons for, and the semantic types of information involved in, each change.

Results: In 7 articles, we identified 206 changes that occurred for 5 fundamental reasons: too much information (29.6%), not enough information (46.1%), inaccuracies (6.8%), misplaced information (14.6%), and structural errors (2.9%). We identified several subcategories; for example, "too much information" was either too detailed, redundant, or extraneous. Changes occurred in all sections of the articles: abstract (15.0%), introduction (4.4%), methods (20.4%), results (29.6%), discussion (26.2%), appendix (2.4%), and moves between sections (2.0%). Individual reasons for change were selectively associated with certain sections and semantic types of information. For example, too little information accounted for 28 (50.0%) of 56 changes to the discussion section. Limitations were frequently omitted.

Conclusions: As result of editorial and peer review, we observed substantial changes throughout all sections of submitted research articles. We propose an analytical framework for categorizing the reasons for these changes. This framework provides a foundation for comparing quantitatively the editorial and peer-review processes across authors, journals, and types of papers. We are confirming the findings of this pilot study in a larger sample of manuscripts.

Return to Session Information