Assessing publication bias in meta-analysis

• Published in: Nursing research (New York) Vol. 52; no. 1; p. 57
• Main Authors: Soeken, Karen L, Sripusanapan, Acharaporn
• Format: Journal Article
• Language: English
• Published: United States 01.01.2003
• Subjects: Humans; Meta-Analysis as Topic; Nursing Research - methods; Nursing Research - standards; Publication Bias - statistics & numerical data; Statistics as Topic - methods
• ISSN: 0029-6562
• DOI: 10.1097/00006199-200301000-00009

Although publication bias is a concern when using meta-analysis, techniques for assessing for its potential appear to be underused in nursing. To demonstrate several methods, both graphical and statistical, of assessing for publication bias. Four methods are described and demonstrated, ranging from the familiar fail-safe N to the trim-and-fill method, using the same contrived data set. Methods vary in their conclusion about the presence of potential publication bias in the example presented. Some limitations of the methods are noted to assist meta-analysts in determining which to consider using. Given that publication bias can affect the validity of meta-analytic studies, researchers are encouraged to use more than one method to assess for bias so that clinicians and decision-makers can have the best possible information available.