Identifying minimally acceptable interpretive performance criteria for screening mammography

Area of Science:

• Radiology
• Medical Imaging
• Quality Improvement

Background:

• Screening mammography interpretation is crucial for early breast cancer detection.
• Establishing clear performance benchmarks is essential for maintaining high-quality diagnostic services.
• Variability in interpretation performance can impact patient outcomes and healthcare resource utilization.

Purpose of the Study:

• To define objective criteria for minimally acceptable physician performance in interpreting screening mammograms.
• To assess the potential impact of implementing these performance criteria on radiology practices in the U.S.

Main Methods:

• Utilized an Angoff approach with expert radiologists to establish performance cut points for sensitivity, specificity, recall rate, and positive predictive values (PPV).
• Phased approach involved expert consensus on performance thresholds and simulation to estimate practice impact.
• Performance measures included sensitivity, specificity, recall rate, PPV(1), PPV(2), and cancer detection rate.

Main Results:

• Defined specific cut points for minimally acceptable performance (e.g., sensitivity <75%, specificity <88% or >95%).
• Estimated that 18%-49% of interpreting physicians might fall below these thresholds, potentially requiring additional training.
• Projected that improvement to acceptable levels could lead to 14 additional cancers detected and 880 fewer false positives per 100,000 women screened.

Conclusions:

• Successfully identified minimally acceptable performance levels for screening mammography interpreters.
• Physicians performing below established cut points warrant review and consideration for further training.
• Implementation of these criteria can enhance the quality and effectiveness of mammography screening programs.