Common Statistical and Research Design Problems in Manuscripts Submitted to High-Impact Public Health Journals

Introduction: Journal editors and statistical reviewers are often in the difficult position of catching statistical and research design problems after data have been collected and analyzed. The authors sought to learn from editors and reviewers of major public health journals what common statistical and design problems they find in submitted manuscripts and what they wished to communicate to authors regarding these issues. Materials and Methodology: Editors and statistical reviewers of 55 high-impact public health journals were surveyed to determine what statistical or design problems they encounter most often. The authors analyzed text responses using content analysis to identify major themes. Results: Editors and reviewers (n = 25) who handle manuscripts from 26 high impact public health journals responded to the survey. The most commonly cited problems included failure to adequately describe statistical models and map them onto research questions, inadequate consideration of sample size, poor control of confounding, and inappropriate reliance on parametric tests. Conclusions: The scientific quality of public health research and submitted reports could be greatly improved if researchers addressed frequently encountered methodological and analytic issues.