Memory bias in observer-performance literature.

Area of Science:

• Medical Imaging
• Observer Performance Studies
• Research Methodology

Background:

• Memory bias can significantly impact the results of observer performance studies.
• Standardized methods for mitigating memory bias in study design are crucial for reliable findings.

Purpose of the Study:

• To evaluate how authors of published observer-performance experiments addressed memory bias in their study designs.
• To identify common practices and potential deficiencies in managing memory bias in diagnostic imaging research.

Main Methods:

• Systematic review of observer performance experiments published in American Journal of Roentgenology and Radiology from 1970 onwards.
• Analysis of 110 selected studies, recording methods for ordering of conditions and images, and use/duration of time gaps between viewings.

Main Results:

• Forty-five of 110 experiments (40.9%) employed methods not susceptible to memory bias.
• Among the remaining 68 experiments, 22.1% used ordering methods that may introduce memory bias, and 16.2% intermixed conditions.
• Reporting completeness regarding memory bias mitigation was lacking in 52.9% of studies.

Conclusions:

• A notable proportion of observer performance studies could enhance their methodologies for ordering tested conditions to minimize memory bias.
• Improved and detailed reporting on methods for ameliorating memory bias is recommended to increase the transparency and validity of research findings.