Related Experiment Videos
Recall and detection rates in screening mammography
David Gur1, Jules H Sumkin, Lara A Hardesty
1Department of Radiology, University of Pittsburgh and Magee-Womens Hospital, Pittsburgh, Pennsylvania 15213, USA. gurd@msx.upmc.edu
Cancer
|April 10, 2004
Summary
Experienced radiologists showed a significant correlation between recall and detection rates in screening mammography. This relationship persisted even after accounting for the volume of mammograms interpreted by each radiologist.
Area of Science:
- Radiology
- Medical Imaging
- Oncology
Background:
- Investigated the correlation between recall and detection rates in screening mammography.
- Focused on 10 experienced radiologists from an academic institution.
Purpose of the Study:
- To determine the relationship between recall rates and tumor detection rates.
- To assess the impact of mammogram interpretation volume on this relationship.
Main Methods:
- Analyzed practice-related and outcome-related databases for 98,668 screening mammograms (2000-2002).
- Calculated recall and tumor detection rates for 10 Mammography Quality Standard Act (MQSA)-certified radiologists.
- Employed Pearson (r) and Spearman (rho) correlation coefficients.
- Utilized partial correlations to control for total reading volumes.
Main Results:
- Observed significant variability in recall rates (7.7-17.2%) and detection rates (2.6-5.4 per 1000 mammograms).
- Found a statistically significant positive correlation between recall and detection rates (P < 0.05).
- This correlation remained significant after controlling for individual radiologist reading volumes.
Conclusions:
- Optimal screening mammography performance requires quantitative evaluation.
- Careful assessment of practice guidelines aimed at reducing recall rates is necessary.
- Individual radiologist performance metrics, including recall and detection rates, are important.