Confidence Interval Calculation Methods Are Infrequently Reported in Emergency-medicine Literature

Abstract

Background: There are many different confidence interval calculation methods, each providing different as well as in some cases inadequate interval estimates. Readers who know which method is used are better able to understand potentially significant limitations in study reports. Objectives: To quantify how often confidence interval calculation methods are disclosed by authors in four peer-reviewed North American emergency-medicine journals. Methods: The authors independently performed searches of four journals for all studies in which comparisons were made between means, medians, proportions, odds ratios, or relative risks. Case reports, editorials, subject reviews, and letters were excluded. Using a standardized abstraction form developed on a spreadsheet, the authors evaluated each article for the reporting of confidence intervals and evaluated the description of methodology used to calculate the confidence intervals. Results: A total of 212 articles met the inclusion criteria. Confidence intervals were reported in 123 articles (58%; 95% CI = 51% to 64%); of these, a description of methodology was reported in 12 (9.8%; 95% CI = 5.7% to 16%). Conclusions: Confidence interval methods of calculation are disclosed infrequently in emergency medicine literature. © 2007 Society for Academic Emergency Medicine.