Confidence ratings and the visual search pattern of expert radiologists appear to be affected by changing prevalence expectations, particularly for normal images, according to a study published in the April issue of Academic Radiology.
It has been recognized that prevalence’s effects on the behavior of radiologists are not well understood. “The prevalence phenomenon is reported to alter radiologists’ behavior and this may be because of the actual differences in prevalence levels or in the expectation of the prevalence level (even when the actual normal-to-abnormal ratio remains constant),” wrote lead author Warren M. Reed, PhD, of the University of Sydney in Australia, and colleagues.
Reed et al sought to better understand the impact of prevalence expectation on radiologists’ behavior by measuring its effect on the performance of 22 experienced radiologists during image interpretation of pulmonary lesions on chest radiographs. The radiologists were allocated into three separate groups, each interpreting a set of 30 images, 15 of which were abnormal. However, prior to the viewings, some radiologists were informed of how many abnormal images to expect in each group, with group 1 being told nine and group 2 being told 22. Group 3 was not told how many abnormalities to expect. The researchers then compared the eye position metrics and receiver operating characteristics confidence ratings between the normal and abnormal images.
The study’s results indicated that significant increases were observed for the duration of image scrutiny and the number of fixations per image in normal images with a higher prevalence expectation. The first group exhibited a significant increase in confidence ratings when prevalence expectation increased for normal images.
The authors also noticed significant increases in duration of image scrutiny in the first group and the number of fixations per image in groups one and two for abnormal images.
“The clinical implications of our findings suggest that changes in explicit prevalence expectation can impact on aspects of readers' behavior,” wrote Reed and colleagues. “The impact of the ‘prevalence effect’ with regards to explicit prevalence expectation is small but real and expert radiologists tend to vary their behavior to a greater extent when interpreting normal images. The impact of prevalence in a variety of radiologic situations requires further attention by researchers and clinicians to assess the potential effects in the clinical setting.”