How to identify information bias due to self-reporting in epidemiological research

Abstract

Reality can be distorted in many ways when seen through a questionnaire or an interview. Such distortions may be systematic, introducing bias. Bias can spoil research by indicating false associations or failing to detect true relationships. It is practically impossible to eliminate measurement errors totally, but estimating the extent of disagreement and assessing whether the errors are systematic should be a priority in epidemiological research. The aim of this article is pedagogically oriented. Through Medline searches and cross-references, 1400 articles were identified, of which 53 were chosen. This review gives an overview of information bias, focusing on recall period, selective recall, social desirability, interview situation and interviewing tools, question phrasing, alternative answers and digit preference. We use a problem identification approach and also present some possible solutions, exemplifying the different topics by research conducted in the fields of HIV-AIDS, nutrition and alcohol abuse. Methods for measuring bias are presented.