Background: The present study was undertaken to ascertain the prevalence of published data with errors in the numerical significant figures in established surgical and medical journals in 2017. The frequency of errors was not only summarized but was also correlated to the published journal impact factor for the seven journals reviewed.
Methods: All original investigations and other analysis reporting quantitative statistical results published in seven surgical and medical journals in 2017 were electronically reviewed for errors in reporting significant figures of the published statistical findings. Errors in significant figures were placed into one of three author defined categories: calculated significant figure errors, interval precision errors, and P value reporting errors. Tests for intraobserver and interobserver reproducibility were conducted blindly to ensure validity and reproducibility between different readers.
Results: A total of 1675 articles published in 2017 were identified and reviewed. In total, 730 articles (44%) were reported to have an error in one category, with error rates ranging from 25% to 68% depending on publishing journal. The error rate for each journal were easily reproduced by different observers (κ coefficient range: 0.55-0.81) and correlated with its 2016 impact factor (r = 0.97, R2 = 0.95, P < 0.001).
Conclusions: Published findings are frequently reported incorrectly in the surgical and medical literature and can be potentially misleading. The pervasiveness of errors correlates to fewer citations as measured by the lower impact factor.
Keywords: Confidence interval; Impact factor; P value; Significant figure; Statistical reporting; error.
Published by Elsevier Inc.