Routine receiver operating characteristic analysis in mammography as a measure of radiologists' performance

Area of Science:

• Radiology
• Medical Imaging Analysis
• Quality Improvement in Healthcare

Background:

• Routine audit data in mammography departments can be leveraged for performance evaluation.
• Assessing radiologist performance is crucial for maintaining diagnostic accuracy and patient care.
• Receiver Operating Characteristic (ROC) curve analysis offers a quantitative method for evaluating diagnostic tests.

Purpose of the Study:

• To describe a system for calculating ROC curves from routine mammography audit data.
• To compare radiologist performance between symptomatic and screening mammography settings.
• To identify specific clinical signs that may pose diagnostic challenges for radiologists.

Main Methods:

• Development of a system to calculate ROC curves using diagnostic opinion and pathology outcome data.
• Collection of data from both symptomatic mammography and breast screening departments.
• Analysis of ROC curves to compare radiologist performance and identify areas of difficulty.

Main Results:

• Radiologists demonstrated better performance in the breast screening centre compared to the symptomatic department.
• Performance differences were not solely due to patient age distribution, suggesting task and population variations.
• ROC curves identified specific clinical signs associated with radiologist difficulty.

Conclusions:

• ROC analysis integrated into routine audits can maintain and enhance individual radiologist performance.
• This method aids in targeting educational efforts towards specific areas of weakness.
• The system provides a valuable tool for quality assurance in mammography interpretation.